|Home | About | Journals | Submit | Contact Us | Français|
Elevated left ventricular filling pressure (LVFP) is a cardinal feature of heart failure with preserved ejection fraction (HFpEF). Mitral E/e’ ratio has been proposed as a non-invasive measure of LVFP. We studied the accuracy of E/e’ to estimate and track changes of LVFP in patients with unexplained dyspnea.
We performed supine and upright transthoracic echocardiography in 118 patients with unexplained dyspnea who underwent right heart catheterization. Supine E/e’ ratio modestly but significantly correlated with supine pulmonary arterial wedge pressure (PAWP) (r=0.36; p<0.001) and demonstrated poor agreement with PAWP values (Bland-Altman limits of agreement of −8.3 to 8.3 mmHg; range: 6.5 to 21.2 mmHg). Similarly, E/e’ ratio cut-off of 13 performed poorly in identifying patients with elevated LVFP (sensitivity 6%, specificity 90%). The ROC area of E/e’ was 0.65 (95% CI: 0.50-0.79). With change from the supine to upright position, PAWP decreased (−5±4 mmHg; p<0.001) as did both E-wave (−17±15 cm/s; p<0.001) and e’ (−2.7±2.7 cm/s; p<0.001) velocities, while E/e’ remained stable (+0.2±2.6; p=0.57). Positional change in PAWP correlated modestly with change in E-wave (r=0.37; p<0.001) velocity. There was no appreciable relationship between change in PAWP and change in average E/e’ (r=−0.04; p=0.77) and in half the patients the change in PAWP and E/e’ were directionally opposite.
In patients with unexplained dyspnea, E/e’ ratio neither accurately estimates PAWP nor identifies patients with elevated PAWP consistent with HFpEF. Positional changes in E/e’ ratio do not reflect changes in PAWP.
Dyspnea is a common symptom, affecting up to half of patients in the inpatient setting and one quarter of patients in the ambulatory setting1. Dyspnea may be caused by cardiac, pulmonary, hematologic, neuropsychiatric or neuromuscular diseases. Among patients with dyspnea and normal left ventricular ejection fraction (LVEF), pulmonary function, and oxygen carrying capacity, the diagnosis of heart failure with preserved ejection fraction (HFpEF) is particularly challenging. Because cardiac structural and functional changes in HFpEF are more subtle and less specific than in heart failure with reduced LVEF, the demonstration of increased left ventricular filling pressure (LVFP) is central to establishing this diagnosis2. The invasive assessment of LVFP is usually done during right heart catheterization (RHC) by measurement of pulmonary arterial wedge pressure (PAWP). Transthoracic echocardiography can be used to non-invasively estimate LVFP. Several echocardiographic measures have been used3, but the ratio of the peak early mitral inflow velocity (E) over the early diastolic mitral annular velocity (e’) has gained wide acceptability in routine clinical practice4. Although integration of multiple echocardiographic measures is always preferred, the E/e’ ratio is often used by itself to appraise LVFP due to the relative ease of acquisition, calculation, and interpretation. The E/e’ ratio has been endorsed by professional guidelines as a surrogate parameter of LVFP in the diagnostic work-up of HFpEF5. In theory, a decrease in LV preload (a major determinant of LVFP) will result in a lower atrial-to-LV diastolic pressure gradient and lower E wave velocity, without significantly affecting e’ which is considered to be relatively independent of pressure-flow gradients6. Despite its widespread use, conflicting data have been reported regarding its accuracy and its ability to track changes in LVFP 4, 7-10. Therefore, we aimed to study the accuracy of E/e’ to estimate and track changes in LVFP, as well as to identify elevated LVFP consistent with HFpEF in patients with unexplained dyspnea and a preserved LVEF.
We studied sequential patients with dyspnea of indeterminate cause referred to the Dyspnea Clinic at Brigham and Women's Hospital between May 2013 and June 2014. All patients underwent resting supine and upright invasive hemodynamic evaluations. Patients with LVEF < 50%, more than mild valvular disease, prosthetic mitral valve and atrial fibrillation were excluded from this analysis. The Partners Human Research Committee approved this study and waived the requirement for informed consent.
A flow-directed, balloon-tipped, 4-port pacing pulmonary artery catheter (Edwards Lifesciences, Irvine, CA) was placed into the pulmonary artery, with ultrasound and fluoroscopic guidance. A second catheter was inserted into the radial artery using a 20-gauge IV or 5-French sheath. End expiratory systemic arterial, right atrial (RAP), right ventricular (RVP), pulmonary artery pressures (PAP) and PAWP were measured using a hemodynamic monitoring system (Xper Cardio Physiomonitoring System, Philips, Andover, Massachusetts) calibrated before each study. The pressure transducer was leveled using as references the mid axillary line (supine) and 5cm below the axillary fold (upright). Cardiac output (CO) was determined by assumed Fick or thermodilution methods during supine RHC, and true Fick method with direct measurement of VO2, arterial and mixed venous O2 content, during upright assessments.
Supine transthoracic echocardiography was performed before patients underwent RHC (time interval <1 hour). After catheterization, upright transthoracic echocardiography was performed with the patient seated resting on the cycle ergometer simultaneously with the invasive hemodynamic measurements. All quantitative echocardiographic measurements were performed by a single reader (M.S.) blinded to invasive hemodynamic data, using a computerized off-line analysis station as previously described11. Peak early diastolic tissue velocity (e’) was measured at the septal and lateral mitral annulus. Mitral inflow velocity was assessed by pulsed wave Doppler from the apical 4-chamber view, positioning the sample volume at the tip of the mitral leaflets. Deceleration time of the E-wave was measured as the interval from the peak E-wave to its extrapolation to the baseline. E/e’ ratio was calculated as E-wave divided by e’ velocities. LV mass was estimated from LV linear dimensions and indexed to body surface area as recommended by ASE guidelines12. LV hypertrophy (LVH) was defined as LV mass indexed to body surface area (LV mass index; LVMi) >115 g/m2 in men or >95 g/m2 in women. LV volumes were estimated by the modified Simpson method using the apical 4- and 2-chamber views, and LVEF was derived from volumes in the standard manner. LA volume was estimated by the method of disks using apical 4- and 2-chamber views at an end-systolic frame preceding mitral valve opening and was indexed to body surface area to derive LA volume index. Measurements were performed in triplicate and the average value used for analysis. Intraobserver reproducibility was assessed in 20 studies randomly selected with the following results. E wave: correlation coefficient 0.99, coefficient of variation 4%; septal e’: correlation coefficient 97%, coefficient of variation 4%; lateral e’: correlation coefficient 98%, coefficient of variation 5%. For the purposes of generalizability, interobserver reproducibility was assessed in the same 20 studies with the following results. E wave: correlation coefficient 0.94, coefficient of variation: 6%; septal e’: correlation coefficient: 0.98, coefficient of variation 7%; lateral e: correlation coefficient 97%, coefficient of variation: 8%.
Continuous variables are expressed as mean ± standard deviation for normally distributed variables or median and interquartile range for non-normally distributed variables. Categorical variables are expressed as number of subjects and proportion [n (%)]. Comparisons between groups were performed using 2-sided parametric or non-parametric tests (unpaired or paired t or Wilcoxon rank sums) for normally and non-normally distributed data respectively. Fisher's exact test was used to compare proportions. Univariate linear regression analysis was performed to model an equation of the relationship between E/e’ ratio and PAWP. Univariate logistic regression was used to study the association between E/e’ ratio the dichotomous variable of elevated LVFP (PAWP ≥ 15 mmHg). For each analysis, separate univariate regression models were generated each of the following predictor variables: lateral E/e’ ratio, septal E/e’ ratio, average E/e’ ratio. To further assess the potential for E/e’ ratio to identify elevated LVFP, receiver operating characteristic (ROC) curve analyses were performed. Correlations between those variables were determined using Pearson or Spearman correlation, as appropriate. Bland-Altman analysis was used to assess agreement between non-invasive and invasive variables. Two supplemental analyses were performed. To assess the impact of missing data on our findings, were performed all analyses restricted to the population of patients with complete data for supine and upright PAWP and E/e’ ratio. In a second supplemental analysis, to assess the potential impact of an outlier value, we repeated the analysis excluding this outlier value. A two sided p-value <0.05 was considered significant. Statistical analysis was performed using Stata software Version 12.1 (Stata Corp LP, College Station, TX, USA).
Of a total of 140 subjects, we excluded 22 because of: LVEF < 50% (n=3), more than mild valvular disease (n=13), valvular prosthesis (n=4) and atrial fibrillation (n=2). Of the remaining 118 patients included in this analysis, 70% were female and the median age was 57 years (1st-3rd quartile: 40-70 years; Table 1). Mean BMI was 27.5±6.5 Kg/m2, and 25% of patients were obese (defined as BMI ≥ 30 kg/m2). The most prevalent comorbidity was hypertension (46%). Average LVMi was 64±19 g/m2, and LVH was present in 23% of patients. LVEF was normal (63±8%) and supine invasive hemodynamics revealed normal average cardiac index (3.0±0.6 L/min/m2), with a mean PAWP of 12±5 mmHg. 26 (22%) had a supine PAWP > 15 mmHg. Echocardiographic data are summarized in Tables 1 and and2.2. Mean E/A ratio was 1.2±0.5, and e’ septal (8.7±2.8 cm/s) was lower than e’ lateral (11.9±4.3 cm/s; p<0.001).
PAWP modestly correlated with E/e’ septal (r=0.41; p<0.001), lateral (r=0.30; p<0.001) and average (r=0.36; p<0.001) (Table 3; Figure 1). E/e’ refers to average E/e’ throughout the remainder of this manuscript unless specifically noted. E/A ratio was even more modestly correlated with PAWP (r=0.21; p=0.04); no correlation was found between DT and PAWP (r=−0.06; p=0.57). Using supine average E/e’ to predict supine PAWP, the linear regression model was: PAWP = 0.44*E/e’ average + 7.2 (N= 88; for the slope: p=0.001, 95%CI for β coefficient for E/e’ was: 0.20-0.68). We computed the predicted PAWP estimated by this linear regression equation, and used the Bland-Altman method to quantify the agreement of the predicted PAWP with the invasively measured PAWP. There was no bias (mean 0 mmHg, 95%CI: −0.8; 0.8), and the limits of agreement were wide (−7.7 to 7.7 mmHg) – Figure 1.
Twenty-two percent (n=26) of the study cohort has a supine RHC PAWP > 15 mmHg13. In a logistic regression model, average E/e’ ratio was not significantly predictive of elevated LVFP (OR=1.09; p=0.22, 95%CI: 0.95-1.27), with a ROC area of 0.65 (95%CI: 0.50-0.79). While the mean PAWP was higher among those with an average E/e’ ≥13 compared to <13 (14±6 vs 11±4 mmHg respectively, p=0.001), an E/e’ average ≥ 13 had a sensitivity of 6% and a specificity of 90% to identify elevated LVFP. Similarly, the ROC areas for E/e’ septal and lateral were 0.67 (95%CI: 0.53-0.81) and 0.62 (95% CI: 0.46-0.78) respectively. An E/e’ septal ≥ 15 had a sensitivity of 6% and a specificity of 92%; an E/e’ lateral ≥ 12 had a sensitivity of 13% and a specificity of 92% to discriminate patients with elevated LVFP.
There was a decrease in cardiac index (−0.3±0.9 L/min/m2) and PAWP (−5±4 mmHg) from supine to upright position (Table 2). Mean arterial pressure increased modestly (+4±12 mmHg; p<0.001), as did systemic vascular resistance (+372±585 dyne.s.cm−5; p<0.001) and heart rate (+8±12 bpm; p<0.001). Both LV diastolic and systolic volumes decreased, with a mild decrease in LVEF (Table 2).
Despite the significant decrease of PAWP, no significant differences were found between supine and upright E/e’ ratio (Table 2). We found no correlation between change in E/e’ (septal, lateral or average) and PAWP (Table 3; Figure 2). In fact, 46% (23/50) of patients had a directionally discordant change of E/e’ when compared to PAWP (Table 4). Those with concordant E/e’ and PAWP with positional change were older (62±16 vs 45±20 years-old; p=0.01), more likely to have hypertension (63 vs 22 %; p=0.003), and had lower myocardial relaxation velocities (e’ average: 8.2±2.5 vs 12.0±3.8 cm/s; p<0.001) and larger left atria (22±7 vs 17±4 mL/m2; p=0.007) than those with discordant changes.
With positional change, significant reductions in E-wave (−17±15 cm/s; p<0.001), e’ septal (−1.8±2.3; p<0.001), and e’ lateral (−3.7±3.5; p<0.001) were noted. The change in E wave velocity significantly correlated with the observed change in PAWP (r=0.41; p<0.001). Change in PAWP was not significantly associated with the observed change in e’ septal (r=0.18; p=0.12), change in e’ lateral (r=0.18; p=0.18), and change in e’ average (r=0.21; p=0.11) – Figure 2.
Supine E/e’ ratio had a significant, though modest, correlation with PAWP, but demonstrated poor accuracy in estimating PAWP evidenced by the wide limits of agreement in Bland-Altman analysis. Concordantly, the recommended E/e’ ratio cut-offs performed poorly in identifying elevated LVFP. Furthermore, change in E/e’ ratio from supine to upright position did not reliably track changes in PAWP, as both E and e’ were significantly affected by hemodynamic changes. In fact, half of the studied patients had a directionally opposite change in E/e’ compared to PAWP, indicating the erratic response of this non-invasive index to load changes associated with the upright position.
Although several previous studies have compared non-invasive (echocardiography-based) estimates of LVFP with invasive data (RHC), our study is one of the largest and – to our knowledge – one of the first to investigate this question in a population of patients with unexplained dyspnea, in whom assessment of LVFP is particularly important for diagnostic purposes. Consistent with previous studies using this methodology, we found a significant correlation between supine E/e’ ratio and PAWP, 14, 15 with a comparable predictive equation based on linear regression modeling 14, 16, 17. However, correlative measurements provide little information about the agreement between supine E/e’ ratio and PAWP. Using Bland-Altman analysis, we found wide limits of agreement between predicted PAWP based on the supine E/e’ ratio and invasively measured PAWP. This indicates a large, clinically significant difference in LVFP estimation when done by these 2 methods. In addition, E/e’ ratio did not accurately identify patients with PAWP > 15 mmHg, and the recommended E/e’ cut-offs had a very low sensitivity to identify elevated LVFP (Figure 3).
Change in E/e’ ratio did not correlate with change in PAWP. The mean E/e’ ratio was higher on average in the upright position compared to supine, while PAWP was significantly lower. This is concordant with a previous study18 and was explained by a more pronounced decrease of e’ than E-wave from supine to upright. The 17% decrease in E-wave velocity reflects a reduced transmitral pressure gradient, which may result from either a lower left atrial pressure and/or a higher proto-diastolic LV pressure. The e’ average demonstrated a 28% decrease with position change from supine to upright, with comparable reductions in both e’ septal and e’ lateral. Together with previous studies19-21 that used different interventions to induce changes in LV preload, these findings clearly demonstrate that e’ is not load-independent. This preload dependence may be more pronounced in patients with a compliant myocardium, which may be more susceptible to changes in external load than a stiff myocardium22. There are several mechanisms that may account for this. First, reduced LV preload due to decreased venous return may influence e’ by decreasing ventricular filling during proto-diastole (rapid filling phase)23. Decreased LV preload can also result in reduced LV systolic torsion, leading to less energy release during early diastolic elastic recoil24.
Second, the increased SVR is associated with an increase of LV afterload that can influence the myocardial relaxation25. However, given the minor changes in systemic vascular resistance and mean arterial pressure relative to the preload changes, and the known LV diastolic tolerance to afterload when LVEF is preserved26, the contribution of increased afterload to the observed positional changes in e’ appears limited.
The analysis of more than one pair of measurements from each patient allows a better understanding of the individual utility of E/e’ ratio, which is often overlooked when only group-derived variables are calculated from measurements at a single time-point. Notably, we observed that almost half (46%) of patients had a discordant E/e’ ratio change compared to PAWP. This expresses a wide inter-individual variation in the way E/e’ ratio relates with PAWP. Some previous studies have reported a significant correlation between change in E/e’ and change in PAWP in decompensated HF, heart transplant and hypertrophic cardiomyopathy patients14, 15, 17, 27, although this finding has been inconsistent. Bhella et al.16 actively manipulated LVFP of HF patients and healthy controls. The authors described the inconsistent relation between group- and individual-derived linear regression slopes. Likewise, Mullens et al.7 did not find a correlation between E/e’ ratio and PAWP changes in decompensated HF patients. The reasons for the discrepancy among these studies may relate to differences in patient characteristics as the presence of valvular regurgitation, interventricular dessynchrony, systolic function, and different degrees of diastolic dysfunction. Consistent with previous studies that showed a greater preload dependence of e’ in subjects with less impaired myocardial relaxation28, 29, we observed greater myocardial relaxation velocities in those patients with discordant changes in E/e’ and PCWP. Differences in the study design as the small sample size of the studied groups, potential selection and ascertainment bias may also account for that discrepancy. Together, these findings seriously question the use of this echocardiographic parameter to track LVFP changes.
This study has several limitations. TDI measurements in the upright position were missing in a subset of patients. However, the absence of differences of clinical, echocardiographic and invasive hemodynamics measurements between patients without versus with TDI missing values argues against a systematic bias in the ascertainment of those echocardiographic images (Supplemental material). In addition, sensitivity analyses restricted to participants without missing data for supine or upright PAWP and E/e’ demonstrated consistent results with primary analysis (Supplemental material). While upright image acquisition was simultaneous with invasive measurements, supine TDI parameters were acquired within 1 hour before the PAWP measurement. Although we cannot exclude some error due to the non-simultaneous measurement, the brief time interval makes large changes in measures unlikely. In addition, we found better correlation on the supine than upright measurements. We used change in position (supine to upright) to examine the relationship between E/e’ ratio and LVFP changes. Although this maneuver evokes a complex cardiovascular response, the observed changes in heart rate and LV afterload surrogates suggest that the predominant change was that in LV preload, supporting the extrapolation of our results to LVFP changes occurring in other clinical scenarios. Finally, correlation analyses can suffer from disproportionate weighting of extreme outliers. We repeated our analysis excluding the observed extreme outlier (Supplemental material) and had similar findings to our primary analysis (Supplemental material). The extrapolation of our results to patients with unexplained dyspnea should be cautious, as this is a very heterogeneous population comprising patients presenting a wide spectrum of structural and functional cardiac abnormalities.
Despite these limitations, our study is one of largest with E/e’ ratio and related invasive LVFP measurements. The studied population is heterogeneous and representative of patients to whom LVFP estimation is commonly done in clinical practice to aid their diagnostic work-up. We did paired measurements for each patient which allows us to go beyond the group mean estimates and better understand the individual response of E/e’ ratio to LVFP changes.
In patients with preserved LVEF referred for RHC because of unexplained dyspnea, E/e’ ratio did not accurately estimate PAWP or identify patients with elevated PAWP consistent with HFpEF. Positional changes in PAWP were not related to changes in the E/e’ ratio especially in those with better myocardial relaxation. These results argues against using E/e’ ratio as an estimator of PAWP in patients with unexplained dyspnea and a preserved LVEF.
Sources of Funding
Work for this manuscript was supported by grant HMSP-ICJ/0013/2012 from the Portuguese Foundation for Science and Technology (M.S.), 1K08HL116792-01A1 from the National Institutes of Health (A.M.S.), and 14CRP20380422 from the American Heart Association (A.M.S.).
Dr Shah reports receiving research support from Novartis, Actelion Pharmaceuticals Ltd, and Gilead. The other authors have no disclosures.