Author + information
- Received August 8, 2016
- Revision received October 18, 2016
- Accepted October 20, 2016
- Published online March 15, 2017.
- Muaz M. Abudiab, MD,
- Lakshmi H. Chebrolu, MD,
- Robert C. Schutt, MD,
- Sherif F. Nagueh, MD and
- William A. Zoghbi, MD∗ ()
- ↵∗Address for correspondence:
Dr. William A. Zoghbi, Department of Cardiology, Houston Methodist DeBakey Heart and Vascular Center, 6550 Fannin Street, SM-677, Houston, Texas 77030.
Objectives This study sought to identify Doppler parameters useful for the assessment of left ventricular filling pressure (LVFP) in patients with mitral annular calcification (MAC) and to develop and validate a decision algorithm for assessing LVFP in such patients.
Background Predicting LVFP in the presence of MAC is problematic.
Methods Prospectively, 50 patients with MAC (mean 72 ± 11 years of age) underwent a complete Doppler echocardiographic study and right or left heart catheterization. Standard and nonstandard parameters of ventricular filling and relaxation were correlated with LVFP. Classification and regression tree analysis was used to develop a decision tree for prediction of LVFP. Validation was performed prospectively using an additional cohort with MAC and invasive hemodynamics (n = 21).
Results In the initial study group, 26 patients had mild MAC, and 24 had moderate or severe MAC. Mean LVFP was 17.0 ± 8.1 mm Hg (range 4 to 50 mm Hg). Of the variables tested, the best predictor of LVFP was the ratio of early-to-late diastolic filling velocity (mitral E/A) (r = 0.66; p < 0.001). This finding was observed in subjects with mild as well as moderate-to-severe MAC. Importantly, the ratio of early diastolic filling velocity-to-mitral annulus velocity (E/e′) demonstrated weak correlation (r = 0.42; p = 0.003). A clinical algorithm using mitral E/A and isovolumic relaxation time (IVRT) was associated with good specificity (100%) and positive predictive value (100%), and moderate sensitivity (81%) and negative predictive value (67%) for high LVFP. Validation of the clinical algorithm in a separate prospective cohort yielded a diagnostic accuracy of 94%.
Conclusions The E/e′ ratio should not be used to estimate LVFP in subjects with significant MAC. However, mitral E/A ratio and IVRT are useful predictors of LVFP in this patient population. The proposed decision algorithm combining these Doppler parameters is accurate in estimating LVFP in patients with MAC.
Echocardiography is frequently used for the noninvasive estimation of left ventricular filling pressure (LVFP) in patients with dyspnea or those suspected of having heart failure. Multiple echocardiographic parameters have been validated for the assessment of LVFP and are routinely obtained in clinical practice (1). Among the most commonly used Doppler parameters is the ratio of mitral early diastolic velocity-to-early diastolic mitral annulus velocity (E/e′), which has performed better overall than other Doppler indices (2–4).
Mitral annular calcification (MAC) is frequently observed with increasing age and in the presence of cardiovascular risk factors (5). Despite heightened recognition of MAC and associated clinical features (5), there remains a paucity of data regarding noninvasive estimation of LVFP in affected subjects. Prediction of LVFP in this setting is problematic because annular calcification may interfere with e′ as an index of myocardial relaxation (6,7). Furthermore, mitral E velocity increases with worsening mitral calcification due to relative mitral inflow stenosis (7). The E/e′ ratio may therefore not be accurate as an index of filling pressure in subjects with MAC, and its use is discouraged in current clinical guidelines (1). Whether LVFP can still be evaluated with Doppler echocardiography in the presence of MAC is not known. We therefore undertook this study to identify the relationship between echocardiographic parameters of diastolic function and LVFP measured invasively in patients with various degrees of MAC and to identify the parameters that best related to LVFP.
The Institutional Review Board of Houston Methodist Hospital approved the protocol, and patients provided written, informed consent. Patients undergoing right or left heart catheterization (or who were being monitored with a pulmonary artery catheter) who had any degree of MAC on a screening echocardiogram were prospectively enrolled to undergo comprehensive echocardiography and hemodynamic study (training group). Those with a technically difficult study, mitral valve prosthesis, tachycardia with monophasic mitral inflow, or mechanical circulatory support device were excluded. In the second part of the study, a prospective validation cohort consisted of patients enrolled using the same protocol. Pertinent clinical data were obtained from the patients and their electronic medical records (Table 1).
Echocardiographic studies and measurements
Patients underwent a complete echocardiogram in the left lateral decubitus or supine position, with an ultrasound system equipped with harmonic imaging and tissue Doppler imaging (TDI) capability (1,8,9). From the apical window, a pulsed Doppler sample volume was placed at the mitral annulus as well as at the mitral tips, and 5 to 10 cardiac cycles were recorded. Continuous-wave Doppler was used to measure peak and mean transvalvular mitral gradients. TDI pulsed Doppler at the mitral annulus was obtained at the inferoseptal and anterolateral corners. In addition, TDI pulsed Doppler of the myocardium was performed with the sample volume advanced 2 cm toward the apex in the apical 4-, 2-, and 3-chamber views. Continuous-wave Doppler was used with the Doppler beam between mitral inflow and LV outflow to record isovolumic relaxation time (IVRT) (1).
Measurements were performed off-line (Digisonics, Houston, Texas) by an observer blinded to the invasive hemodynamic findings. Normal systolic function was defined as left ventricular ejection fraction (LVEF) ≥50%. The mitral inflow velocity was traced, and the following variables were obtained: peak velocity of early (E) and late (A) filling, and deceleration time (DT) of the E-wave velocity. Duration of mitral A-wave was measured at the level of the annulus. IVRT was measured, and the peak and mean transmitral diastolic pressure gradients were calculated. All velocities were averaged over at least 3 cardiac cycles. Too few studies had technically adequate pulmonary vein signals to allow inclusion.
Presence of MAC was assessed in the parasternal short-axis view at the level of the mitral annulus (Figure 1). MAC was graded as mild (focal echodensity with less than one-third involvement of the ring circumference), moderate (involvement of greater than one-third and less than one-half of the ring circumference), or severe (involvement of greater than one-half of the ring circumference or intruding into the LV inflow tract) (10). Interobserver variability in grading MAC severity was assessed in 17 cases (6 mild, 5 moderate, and 6 severe).
Hemodynamic data were obtained at end expiration by an investigator blinded to the echocardiographic measurements. Pressures were collected after the transducers were leveled at the mid-axillary line. LVFP was measured as either pulmonary capillary wedge pressure during right heart catheterization or pre-A LV diastolic pressure during left heart catheterization as previously described (3,11). Data were averaged over 3 to 5 cardiac cycles. Echocardiographic and hemodynamic studies were acquired simultaneously in the majority of participants (62%), and all within 4 h and with no intervening clinical change or therapeutic intervention.
Continuous data are mean ± SD. Regression analysis was used to relate echocardiographic and Doppler measurements to LVFP. Receiver-operating characteristic (ROC) curves were used to identify threshold values that separated patients with low (≤12 mm Hg) from high (>12 mm Hg) LVFP. Classification and regression tree analysis was used to develop a decision-making algorithm. Inter-rater agreement for evaluation of the severity of MAC classified as mild, moderate, or severe was assessed using a linear weighted Cohen kappa statistic. The interclass correlation coefficient was used to assess interobserver and intraobserver agreement of Doppler data in 20 cases. A 2-sided p value <0.05 was considered significant. All statistical analyses were completed using R for Statistical Computing version 3.1.0 software (R Foundation, Vienna, Austria).
The final population consisted of 71 patients (50 in the training group, 21 in the validation group). Clinical characteristics and echocardiographic variables of the training group are presented in Tables 1 to 3⇓⇓. There were 26 subjects with mild MAC, and 24 with moderate or severe MAC. Subjects were at advanced age (range 48 to 92 years old), predominantly male, with a high incidence of overweight, systemic hypertension, and chronic kidney disease. No subjects had severe mitral regurgitation.
There was a wide range of LVEF (23% to 89%), but the majority of patients (74%) had normal systolic function. Left atrial enlargement was common. Mitral E velocity ranged between 47 and 191 cm/s, and average e′ ranged between 2 and 13 cm/s. Compared with patients with mild MAC, those with moderate or severe MAC had higher mitral E and E/e′.
Relationship between Doppler echocardiographic variables and LVFP
Left ventricular filling pressure averaged 17 ± 8 mm Hg (range 4 to 50 mm Hg). Several echocardiographic and Doppler parameters were related to LVFP (Table 4). The parameter that best correlated with LVFP was the conventional mitral E/A ratio measured at the leaflets tip (r = 0.66; p < 0.001) (Figure 2). This finding was observed in subjects with mild (r = 0.76; p < 0.001) as well as moderate-to-severe MAC (r = 0.59; p = 0.005). Correlation was significant in patients with depressed LVEF (r = 0.88; p < 0.001; LVEF = 36 ± 8%; n = 13) and was lower, but significant, in patients with normal LVEF (r = 0.62; p < 0.001; n = 37). In all subjects, the area under the curve (AUC) for prediction of LVFP of >12 mm Hg with mitral E/A was 0.84. For LVFP >12 mm Hg, a mitral E/A ratio of 0.8 was 97% sensitive and 62% specific. However, a ratio of >1.8 was associated with optimal specificity (100%), but poor sensitivity (34%). In cases of mild MAC, sensitivity was 100% at a ratio of 0.8, whereas specificity was 100% at a ratio of 1.8 (AUC: 0.88). In moderate-to-severe MAC, sensitivity was 94%, and specificity was 100% at ratios of 0.8 and 1.8, respectively (AUC: 0.77).
Other Doppler variables related to LVFP but with weaker correlations. Importantly, the average E/e′ ratio demonstrated weak correlation with LVFP (Figure 2). There were negligible differences in correlation when using inferoseptal versus anterolateral annulus velocity. For all patients, ROC analysis for average E/e′ yielded an AUC of 0.68. An average E/e′ cutoff of 14 yielded optimal sensitivity and specificity for high LVFP of 74% and 67%, respectively. In subjects with moderate or severe MAC, however, sensitivity and specificity were 82% and 20%, respectively, with an AUC of 0.45.
Both DT and IVRT related significantly to LVFP (Table 4). Although DT had slightly better correlation, IVRT demonstrated superior prediction of elevated LVFP (AUC: 0.91). An IVRT threshold of 80 ms was optimal with a sensitivity of 75% and specificity of 100%. Although left atrial volume index demonstrated poor linear correlation, a left atrial volume index cutoff of >34 ml/m2 yielded a sensitivity of 83% for LVFP >12 mm Hg. A pulmonary artery systolic pressure of ≤30 mm Hg was 91% specific for normal LVFP.
E/e′ using various sampling locations
Because MAC tends to involve the posterior annulus, further analysis was conducted using TDI sampling in all annular segments. Velocities were also recorded with the pulsed sample volume advanced further apically from the annulus (∼2 cm) into the LV myocardium in the apical 4-, 2-, and 3-chamber views. The best correlation with LVFP for the total population was seen using E/e′ with e′ taken at the annulus of the anteroseptum (r = 0.58; p < 0.001). However, correlation was poor (r = 0.12; p = 0.68) for subjects with moderate or severe MAC and best for those with mild calcification. An E/e′ cutoff of 15 yielded 65% sensitivity and 88% specificity for high LVFP. The number of cases where e′ at the annulus of the anterior wall was of adequate technical quality was small and precluded statistical analysis. Correlation of E/e′ using e′ taken at the inferior (r = 0.40; p = 0.009) or inferolateral annulus (r = 0.37; p = 0.021) was poor. Overall correlations were better in patients with mild MAC, regardless of the location of the annular site sampled. Measuring e′ further into the LV myocardium for all wall segments did not improve the correlation of E/e′ with LVFP (r range = 0.15 to 0.48).
Relation of mitral valve gradient to inflow velocities
No significant correlation was observed between LVFP and mitral E velocity at valve tip, at the annulus (Eann), or the E/Eann ratio. As expected, mean gradient (MG) was higher in subjects with more MAC. The average MG was 2.1 ± 1.0 mm Hg in patients with mild MAC compared with 3.5 ± 1.7 mm Hg in those with moderate-to-severe MAC. Importantly, in cases of higher MG (>3.5 mm Hg), correlations with LVFP, including that of mitral E/A (r = 0.44; p = 0.20), worsened. Conversely, subjects with moderate-to-severe MAC and MG ≤3.5 mm Hg demonstrated improved correlation of mitral E/A with LVFP (r = 0.75; p = 0.005), whereas E/e′ continued to perform poorly.
Suggested clinical algorithm for estimation of left ventricular filling pressure
Classification and regression tree analysis was used to develop a decision tree for prediction of LVFP. A clinical algorithm using mitral E/A and IVRT (Figure 3) was associated with good specificity (100%) and positive predictive value (100%) for high LVFP with moderate sensitivity (81%) and negative predictive value (67%). The algorithm was successfully applied in 44 of 50 patients (88%) with a diagnostic accuracy of 86%. The subset of patients with intermediate mitral E/A (0.8 to 1.8) and higher IVRT (≥80 ms) yielded a high rate of false negatives (5 of 9 patients or 56%). However, the number of patients in this group was small, and the cases classified incorrectly had borderline LVFP (mean of 16 ± 1.5 mm Hg; range 14 to 18 mm Hg). Inability to apply the algorithm was due to atrial fibrillation (n = 1), fusion of mitral E and A associated with tachycardia (n = 1), or inadequate or unavailable Doppler signals (n = 4).
Validation of clinical algorithm
Prospective validation of the clinical algorithm to estimate LVFP was performed. This group consisted of 21 patients (average 71.0 ± 8.1 years of age) with mild (n = 4) or greater MAC (n = 17) and mean LVEF of 56 ± 21%. Representative examples are shown in Figure 4. Application of the clinical algorithm for prediction of elevated LVFP resulted in sensitivity of 100% and specificity of 90%. The algorithm could be applied in 18 of 21 patients (86%) with a diagnostic accuracy of 94%. Mitral E/A ratio was not available in 3 cases due to atrial fibrillation (n = 1) or tachycardia (n = 2). Applying the algorithm to a merged population containing both training and validation groups resulted in a diagnostic accuracy of 89%. Sensitivity and specificity were 85% and 95%, respectively.
Interobserver and intraobserver agreement
Reproducibility for grading of MAC severity and for measurement of the 2 echocardiographic variables used in the proposed clinical algorithm were as follows: exact agreement in grading MAC was 82%, with no more than 1 grade difference in severity. The inter-rater agreement was very good (κ = 0.80). The intraclass correlation coefficients for interobserver agreement for IVRT and mitral E/A were 0.92 and 0.89, respectively. The intraclass correlation coefficients for intraobserver agreement were 0.85 and 0.80 for IVRT and mitral E/A, respectively. Both the interobserver and intraobserver disagreements within the algorithm resulted in reclassification of 1 case (5%) each.
The current investigation evaluated for the first time the accuracy of Doppler parameters in assessing LVFP in patients with MAC, a frequent and clinically challenging situation. The study confirms that E/e′ is not a good parameter of diastolic function in this setting. This is particularly true in patients with moderate or severe MAC. Among all variables tested, the traditional mitral E/A ratio related best to LVFP, albeit modestly. The best combination of parameters for prediction of LVFP was mitral E/A and IVRT. The proposed algorithm can classify most cases (87%), has excellent specificity for high LVFP, and is simple and obtainable in most patients.
Assessing diastolic function in the presence of annular calcification
Patients with MAC are at increased risk of cardiovascular disease and events (5). Evaluating dyspnea in these patients can be a complex endeavor because they have multiple comorbidities. To date, there have been no studies evaluating the prediction of LVFP in this challenging situation.
Estimation of LVFP in mitral stenosis has previously been studied; however, significant stenosis due to annular calcification, although possible, is uncommon (12,13). Diwan et al. (14) reported a poor correlation of average E/e′ with LVFP in patients with mitral stenosis. There was no correlation with DT, but IVRT showed a moderate correlation, consistent with our findings. Indices that do not incorporate velocity have been proposed in mitral stenosis such as the ratio of IVRT-to-interval between onset of mitral inflow and onset of early diastolic velocity (IVRT/TE–e′) (14) and systolic fraction of pulmonary venous flow (15).
Ratio of E/e′ and LVFP
The poor performance of E/e′ in patients with MAC is expected, because mitral inflow velocities are increased (7), whereas annular velocities are reduced (6,7,16). Our results are also consistent with previous observations that changes in mitral E and e′ are directly proportional to the amount of MAC (7). When segments less commonly affected by MAC (i.e., anteroseptum) were studied or the severity of MAC was mild, correlation improved (6). Although we further hypothesized that sampling lower into the LV myocardium may reduce the impact of MAC by distancing the measurement from the annulus, E/e′ continued to perform poorly.
Physiological rationale for the study findings
In the absence of MAC, impaired LV relaxation with normal left atrial pressure results in a slower rate of LV isovolumic pressure fall, a reduction in mitral E, and more filling during atrial contraction. As left atrial pressure rises, however, the predominance of ventricular filling occurs in early diastole with earlier onset, resulting in a larger mitral E/A ratio and shorter IVRT. This same pathophysiological mechanism appears to apply in subjects with MAC. The degree of obstruction by MAC to mitral inflow likely affects both E and A velocity, preserving the diagnostic validity of a high ratio to predict LVFP (taken as a surrogate of left atrial pressure). Given the demographic profile of subjects with MAC, the burden of diastolic dysfunction is very high, if not ubiquitous. Such patients will therefore have either a grade I (mitral E/A of <0.8) pattern of abnormal relaxation and normal left atrial pressure or higher grade patterns (II or III), consistent with elevated left atrial pressure (1). Our results indicate that the greatest specificity is attained for high LVFP in subjects with a ratio (>1.8), approaching a restrictive pattern. IVRT, which relates to relaxation and inversely to pressure, helped reclassify intermediate mitral E/A ratios (0.8 to 1.8) and identify most patients with high LVFP.
The challenge in enrolling subjects with MAC who require heart catheterization within the same time window hampered our ability to acquire an even larger sample size. There is no universally accepted standard for assessment of MAC severity. Echocardiographic examination in the hospital setting introduced technical difficulty. Therefore, variables such as pulmonary vein Doppler analysis and IVRT/TE-e′ were not assessed. IVRT is not routinely reported; however, it is easily obtainable and reproducible with proper technique. Last, in cases of atrial fibrillation, fusion of mitral E and A, or significant mitral stenosis, the proposed algorithm is not applicable. Caution should be also exercised in patients with MAC with a gradient of >3.5 mm Hg. An alternative approach to integrating multiple echocardiographic variables may be devised depending on desire for sensitivity versus specificity for elevated LVFP. Specifically, both left atrial volume index and pulmonary artery systolic pressure demonstrated excellent sensitivity while IVRT had good specificity for prediction of elevated LVFP.
As the prevalence of MAC continues to rise, the ability to estimate LVFP as a diagnostic and prognostic tool in affected patients becomes increasingly important. The current investigation proposes a Doppler algorithm that may be used to estimate LVFP in most patients with MAC. Although further study is needed, this represents a useful tool in the noninvasive assessment of such patients.
COMPETENCY IN MEDICAL KNOWLEDGE: Because annular calcification affects standard parameters of diastolic function, noninvasive estimation of LVFP in patients with MAC has been challenging.
COMPETENCY IN PATIENT CARE AND PROCEDURAL SKILLS: Physicians interpreting echocardiograms may employ the clinical algorithm presented herein to estimate LVFP in patients with MAC.
TRANSLATIONAL OUTLOOK: Additional studies in larger, more varied populations will lend further credence to the diagnostic approach outlined.
The authors have reported that they have no relationships relevant to the contents of this paper to disclose. Drs. Abudiab and Chebrolu contributed equally to this work. A. Jamil Tajik, MD, served as the Guest Editor for this article.
- Abbreviations and Acronyms
- mitral late diastolic velocity
- area under the curve
- deceleration time
- mitral early diastolic velocity
- early diastolic mitral annulus velocity
- isovolumic relaxation time
- left ventricular ejection fraction
- left ventricle filling pressure
- mitral annular calcification
- mean gradient
- tissue Doppler imaging
- Received August 8, 2016.
- Revision received October 18, 2016.
- Accepted October 20, 2016.
- 2017 American College of Cardiology Foundation
- Nagueh S.F.,
- Smiseth O.A.,
- Appleton C.P.,
- et al.
- Ommen S.R.,
- Nishimura R.A.,
- Appleton C.P.,
- et al.
- Nagueh S.F.,
- Middleton K.J.,
- Kopelen H.A.,
- Zoghbi W.A.,
- Quinones M.A.
- Dokainish H.,
- Zoghbi W.A.,
- Lakkis N.M.,
- et al.
- Abramowitz Y.,
- Jilaihawi H.,
- Chakravarty T.,
- Mack M.J.,
- Makkar R.R.
- Ariza J.,
- Casanova M.A.,
- Esteban F.,
- Ciudad M.M.,
- Trapiello L.,
- Herrera N.
- Rudski L.G.,
- Lai W.W.,
- Afilalo J.,
- et al.
- Jassal D.S.,
- Tam J.W.,
- Bhagirath K.M.,
- et al.
- Nagueh S.F.,
- Lakkis N.M.,
- Middleton K.J.,
- Spencer W.H. III.,
- Zoghbi W.A.,
- Quinones M.A.
- Osterberger L.E.,
- Goldstein S.,
- Khaja F.,
- Lakier J.B.
- Diwan A.,
- McCulloch M.,
- Lawrie G.M.,
- Reardon M.J.,
- Nagueh S.F.