Left ventricular diastolic dysfunction: identifying presence by left atrial function

Objective: The left atrium (LA) is exposed to left ventricular pressure during diastole. Applying the 2016 American Society of Echocardiography left ventricular diastolic function (LVDF) guidelines, this study aims to investigate whether left atrial ejection fraction (LAEF) and left atrial active emptying fraction (LAAEF) are markers of diastolic dysfunction (LVDD). Methods: Retrospective cohort of consecutive patients ( n = 124) who underwent transthoracic echocardiography were studied. Doppler peak velocities of passive (MV E) and active filling (MV A) were measured and ratio E/A calculated. Tissue Doppler imaging parameters of peak early (e ′ ) of the septal and lateral mitral annulus were measured, and average E/e ′ ratio (E/e ′ ) was calculated. Tricuspid regurgitation velocity, left atrial maximum volume, left atrial minimum volume and LA volume pre-contraction were measured, allowing calculation of LAEF and LAAEF. Subjects were assigned LVDF categories. Results: Binomial logistic regression model ( X 2 (2) = 48.924, P < 0.01) determined that LAEF and LAAEF predicted diastolic dysfunction with sensitivity 85.5% and specificity 78%. ROC curves determined good diagnostic accuracy for LAEF and LAAEF to predict LVDD, AUC 0.826 and 0.861 respectively. Logistic regression model ( X 2 (2) = 39.525, P < 0.01) predicted those patients with E/e ′ ≥ 14 using LAEF and LAAEF with sensitivity 51.6% and specificity 92.4%. Moderate correlations were found between E/e ′ and log derivatives of LAEF and LAAEF. Conclusions:


Introduction
The cardiac cycle is a complex process of tri-phasic relaxation/filling and bi-phasic ejection. The left ventricular (LV) diastolic process consists of four phases: isovolumic relaxation, passive early rapid ventricular filling and active atrial contraction. The primary role of the left atrium (LA) is to assist LV filling in three phases. A reservoir for pulmonary venous return (PVr) during ventricular systole; a conduit to the LV for PVr during early ventricular diastole and a pump to augment LV filling in late diastole (1,2). The LA is therefore exposed to LV diastolic pressures via the open mitral valve (MV) throughout most of diastole (3,4) and is susceptible to remodelling through increasing pressure and volume. LA size and volume (indexed to body surface area) is therefore considered a marker for left ventricular diastolic function (LVDF) and is included in the LVDF echocardiographic protocol set out by the American LA kinetic energy (7), LA ejection fraction (LAEF) (8) and left atrial expansion index (8).
It has long been known that as vascular resistance increases through reduced arterial compliance, ventricular output falls (9). By the same principle, if LA afterload increases through reduced LV compliance, LA output and consequently left atrial active emptying fraction (LAAEF) and LAEF may also fall. Although LAEF has been previously researched as a measure of LAf in a small number of studies (1,10,11,12,13,14), Hsaio et al. are the only group to have correlated LAAEF (volume at atrial systole (preAv) minus LA minimum volume (LAmin) divided by preAv), with LVDF (8). Using the new ASE LVDF guidelines (5), this study aims to determine if LAAEF or LAEF correlate with or predict the presence of LVDF and therefore provide additional indicators of LVDF severity.

Study population
This was a retrospective cohort study of patients referred for TTE at Royal Papworth Hospital NHS Foundation Trust prior to January 2016. Patients were undergoing TTE for either: open access referral made by their General Practitioner (GPOA) due to breathlessness or by a cardiothoracic surgeon prior to coronary artery bypass graft (CABG) surgery. Data were stored on the Philips Xcelera system (version 3.2.1.712, Koninklijke Philips Electronic N.V2011). The study aimed to collect the data of 100 subjects spread throughout the LVDF categories, indeterminate cases were excluded from statistical analysis. TTE studies were assessed using a purposive sampling method of consecutive studies analysed in reverse chronological order until each ASE-defined diastolic category had a comparable number of studies ( Fig. 1). Grade III diastolic dysfunction (restrictive filling) category was underpopulated due to lack of patients who fulfilled the inclusion criteria.
The study was approved by Royal Papworth Hospital NHS Foundation Trust Research and Development Department.

Inclusion/exclusion criteria
All subjects referred via GPOA or pre-op assessments routes and who underwent a full TTE at Royal Papworth Hospital NHS Foundation Trust prior to January 2016 were considered for this study. Additionally, inpatients requiring full TTE prior to CABG were also considered. The minimum age was 16 years; there was no maximum age cut-off, as although age impacts LV relaxation, it does not affect interpretation of LVDF.
Subjects were excluded from the study, if at the time of the TTE they were in atrial fibrillation, atrial flutter or sinus tachycardia; had evidence of multiple ectopic beats on saved loops or if image quality was inadequate for accurate measurement. The following exclusion criteria were applied: 1) ≥moderate left-sided valvular heart disease, 2) previous MV surgery, prosthetic valves or annular rings, 3) atrial septal closure device, 4) pericardial effusion, 5) primary pulmonary hypertension (eccentricity index >1), 6) LV outflow tract obstruction (either at rest or with provocation), 7) patients receiving dialysis, haemofiltration or inotrope infusions, 8) prior heart transplantation and 9) history of AF ablation.

LAEF measurements
TTEs performed on the Philips IE33 or CX50 were retrieved from the Philips Xcelera system. All measures of LA volume were performed by British Society of Echocardiography accredited sonographers (15).   Left atrial volume (LAV) was measured at three points in the cardiac cycle, in both apical four chamber (A4C) and apical two chamber views: left atrial maximum volume (LAmax) at end systole, LAmin at end-diastole, frame before MV closure and preAv, frame before MV re-opens for atrial kick.
The parameters used in analyses were:

LVDF measurements
The ASE protocol for measurement of LVDF parameters was applied (5) to archived A4C images of trans-mitral inflow and mitral annular velocities from pulsed-tissue Doppler imaging (TDI). Peak velocities of the early (MV E) and late (MV A) phases of the mitral inflow from the Doppler recordings were measured and their ratio (E/A) calculated. Peak early diastolic velocity (e′) of the septal and lateral mitral annulus was measured. The E/e′ was calculated in each case. Additionally, tricuspid regurgitation velocity and LAmax indexed by body surface area were measured. LVDF category was then determined according to the algorithms within the ASE protocol.

Statistical analysis
Data are presented as mean ± standard deviation for continuous variables. Initially, ordinal logistic regression analysis was run to predict LVDF category according to LAAEF and LAEF. However, assumptions necessary for the test were violated, this test was therefore invalidated. Alternatively, binomial logistic regression analyses were run to determine the sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of predicting LVDF using LAAEF and LAEF. ROC curves were used to determine the strength of the predictive tests; accuracy of parameters for predicting LVDF and to define measurement values producing the highest sensitivity and specificity to predict left ventricular diastolic dysfunction (LVDD). Univariate correlations between log LA variables and E/e′ were assessed using Pearson's correlation co-efficient (r) for normally distributed data. Reproducibility of the measurements was assessed by inter-class correlation coefficients (ICCs). For all statistical analyses, a twotailed P < 0.01 was considered significant. Statistical analyses were performed using SPSS software, V.21.0 (SPSS Inc.).

Reproducibility of LA volumes
Reproducibility of LA volumes was assessed in ten randomly selected subjects. One accredited supervisor re-measured the atrial volumes of the subjects, blinded to the original measurements, ICC = 0.959, P < 0.01. LA volumes for ten additional randomly selected subjects were re-measured by the primary operator, blinded to the initial results, ICC = 0.898.

LA function and LVDF
The relationship between LAAEF and LAEF with LVDF was assessed using the most recent ASE guidelines. Subjects with indeterminate LVDF described by ASE guidelines were removed from the following statistical tests. A binomial logistic regression was performed to ascertain the effects of LAAEF and LAEF on the likelihood that patients have LVDD. The logistic regression model was statistically significant X 2 (2) = 48.924, P < 0.01, of the two predictor variables, LAAEF was statistically significant (P < 0.01). Sensitivity, specificity, PPV and NPV are shown in Table 2. This indicates that 86.8% of patients who have LVDD were also predicted to have LVDD by this model (true positives). Additionally, 76.2% of patients who have normal LVDF were predicted to have normal LVDF by this model (true negatives). ROC curves were performed to determine the diagnostic accuracy of using LAAEF and LAEF to predict LVDD (Fig. 2). AUC 0.861 and 0.826 suggest that LAAEF and LAEF respectively are strong indicators of the presence of LVDD. When considering an optimal balance between maximum sensitivity and specificity, LAAEF ≤38.5% and LAEF ≤52.5% predicts diastolic dysfunction with sensitivity 79.7% and 72.5%, and specificity 82.9% and 73.2% respectively (Table 3).   Bar graphs and box plots display the relationship between LAAEF and LAEF, with LVDF. These demonstrate that as LVDF worsens, LAAEF and LAEF decrease (Figs 3 and 4).

Relationship between average E/e′ and LA function
A binomial logistic regression was performed to ascertain the effects of LAAEF and LAEF on the likelihood that patients have E/e′ ≥14. The logistic regression model was statistically significant X 2 (2) = 39.525, P < 0.01; results are shown in Table 4. Of the two predictor variables none were statistically significant.
Additionally, ROC curves were performed to determine the diagnostic accuracy of LAAEF and LAEF to predict average E/e′ (Fig. 5). AUC 0.828 and 0.846 respectively identify these measures as good indicators of LVDD. Co-ordinates are summarised in Table 5.
Pearson's product-moment correlation was performed to assess the relationship between E/e′ and the natural log of LAAEF and LAEF. Statistically significant moderate correlations were found (Table 6) and are illustrated in scatterplots (Fig. 6).

Relationship between LVDF and LA function
When LA pre-load is stable, indexed LAVs are a robust marker for LV filling pressures (E/e′) (4). In this study, we assessed the relationship between LAAEF and LAEF with grades of LVDF as defined by recent ASE guidelines (5). Our findings demonstrate that both predictor variables are capable of predicting LVDD with high sensitivity and specificity. LAAEF ≤38.5% (sensitivity 79.4% and specificity 82.9%) was found to be more sensitive and specific than LAEF.
Filling of the LA during ventricular systole (atrial reservoir phase) is contributed through contraction of longitudinal fibres within the LV. Shortening of the LV towards the apex and mechanical traction of the mitral annulus stretches the LA, in turn decreasing mean LA pressure and augmenting flow from the pulmonary veins; LAmax is therefore achieved at the end of ventricular systole (16). However, despite being considered a key indicator of LVDF in current guidelines, not only is LAmax measured during ventricular systole, it is also influenced by both the degree of LV longitudinal contraction and the compliance of the LA itself. It is therefore clear that additional measures are required to improve the accuracy of TTE for the diagnosis LVDF. Only one other study was found in the literature that compared Simpsons method of discs derived LAAEF with LVDF (8). Hsiao et al. concluded that LA parameters offer a better estimation of LVDF than TDI and pulsed-wave Doppler measurements (8).
Similar findings to this study were also found by both Russo et al. and Scherr et al. (4,17). Both studies assessed the relationship between LAAEF and LAEF and LVDF by real-time 3D echocardiography (4). Scherr et al. found a stronger relationship between LAAEF and LVDF than the one between LAmin and LVDF.
Teo et al. studied the impact of LVDF on LAf and volume and found that LA active contraction is augmented to compensate for changes in LV diastolic properties (18). However, they described this compensatory mechanism failing as the LA becomes dysfunctional and the severity of LVDD increases, causing a lower total emptying volume (18). This describes how the LA behaves similarly to the LV in regards to the Frank-Starling law. Output will increase with an increase in LA size, which aids maintenance of normal stroke volume. Eventually, the output plateaus with continued increase in LAV, until finally moving onto the descending limb of the Frank-Starling curve when contractile function may decrease in the presence of severe LA dilation (19).
In conclusion, our study has demonstrated that increasing LA afterload, secondary to worsening LVDD,

Limitations
Due to the retrospective nature of data collection, suboptimal image quality had to be accepted in some cases. Due to the size of our study sample and underpopulation of patients with grade III LVDD, results would need to be validated over a greater population. Doppler derived velocities and flow ratios are often influenced by the respiratory cycle. Because respiration alters diaphragmatic traction, and therefore, the position of intra-cardiac structures in relation to a fixed Doppler sample, the angle of insonation between the ultrasound beam and area of interest is variable. This consequently leads to variations in peak velocity measurements and potentially variable parameters of LVDF (20). The phases of respiration also affect TDI in the same manner (20).
There is currently no echocardiographic 'gold standard' for quantifying LV end-diastolic pressure, potentially increasing the overall error and accuracy of this study. However, the measures used are validated against the ASE LVDF algorithm. Nevertheless, a raised average E/e′ is rarely seen in normal individuals, but a normal E/e′ can be seen in those with grade 2 and 3 LVDD. Consequently, this study may have excluded some subjects with grade 2 and 3 LVDD on the basis of normal E/e′ and discordant values; however, subjects with raised E/e′ are very likely to have high LV-filling pressures.

Suggestions for future research
Although further validation is required before LAAEF and LAEF can be considered an additional measurement for the assessment of LVDF, we believe that these measures could be useful parameters in subjects where LVDD cannot be confirmed by current guideline algorithms. In addition, subjects with significant valvular disease and recipients of heart transplants were excluded from this cohort due to the difficulties in assessing LVDF in these groups. In such patients, where LVDD may co-exist or potentially indicate allograft rejection, research into LAAEF and LAEF , correlations with the standard LVDF measurements is recommended.
There is a growing body of evidence to support the estimation of LA strain for the assessment LVDF. However, considering that reference values have not yet been established and that significant expertise is necessary for accurate measurement, LA strain is not yet recommended for use within daily practice. In contrast, estimation of chamber volumes and ejection fraction are familiar concepts and routine practice to all sonographers, they are also supported by recognised reference ranges for LA size. In view of this, this study sought to identify additional parameters of LVDF that could be routinely performed on a daily basis, without the need for extensive further training and experience.

Conclusion
LVDD has not previously been compared with LAAEF and LAEF within the same 2D TTE study, or using the 2016 ASE diastolic function guidelines. The findings of this study clearly show that LAAEF and LAEF are strong predictors of LVDD with high sensitivity and specificity and could be considered valid additional markers to the standard dataset for assessing LVDF; as LVDD worsens, LAAEF and LAEF decrease.

Declaration of interest
The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the research reported.