Left Atrial Size and Heart Failure Hospitalization in Patients with Diastolic Dysfunction and Preserved Ejection Fraction.

CONTEXT: Heart failure with preserved ejection fraction (HFpEF) is a clinical syndrome associated with diastolic function abnormalities. It remains unclear which factors, if any, can predict the transition from asymptomatic diastolic dysfunction to an overt symptomatic phase.
MATERIALS AND METHODS: Patients hospitalized with suspected heart failure between January 2012 and November 2014 with a transthoracic echocardiogram demonstrating preserved systolic function were screened (n = 425). Patients meeting the American College of Cardiology Foundation/American Heart Association definition for HFpEF (n = 40) were matched in a 1:1 fashion to individuals admitted for hypertensive urgency with diastolic dysfunction and neither pulmonary edema nor history of heart failure (n = 40). The clinical records and echocardiograms of all eighty patients included in this retrospective study were reviewed.
RESULTS: Patients with HFpEF had higher body mass index (BMI), creatinine, beta-blocker use, and Grade 2 diastolic dysfunction when compared to the hypertensive control population. Echocardiographic analysis demonstrated higher right ventricular systolic pressures, left ventricular mass index, E/A, and E/e' in patients with HFpEF. Similarly, differences were observed in most left atrial (LA) parameters including larger LA maximum and minimum volume indices, as well as smaller LA-emptying fractions in the heart failure group. Multivariate logistic regression analysis revealed LA minimum volume index (odds ratio [OR]: 1.23 [1.09-1.38], P = 0.001) to have the strongest association with heart failure hospitalization after adjustment for creatinine (OR: 7.09 [1.43-35.07], P = 0.016) and BMI (OR: 1.11 [0.99-1.25], P = 0.074).
CONCLUSION: LA minimum volume index best correlated with HFpEF in this patient cohort with diastolic dysfunction.

Introduction

Heart failure with preserved ejection fraction (HFpEF) is a clinical syndrome that continues to gain recognition for its rising prevalence and rates of morbidity and mortality that mirror the syndrome of heart failure with reduced ejection fraction.[1] Unfortunately, there remains a poor understanding of the underlying pathophysiology in patients with HFpEF, and only a few existing therapeutic strategies have proven to improve outcomes.[1]

HFpEF is defined as typical symptoms and signs of heart failure in a patient with normal left ventricular (LV) ejection fraction, in the absence of significant valvular abnormalities by echocardiogram.[2] The onset of this syndrome is typically preceded by asymptomatic phase of LV diastolic dysfunction that over time can evolve into clinically overt failure, disability, and even death. Thus, the emphasis should be placed on the detection of these abnormalities in the preclinical phase.[34]

In addition to diastolic grading, there are other echocardiographic findings such as LV hypertrophy (LVH) and increased left atrial (LA) size that have been associated with poor cardiac outcomes in patients with HFpEF.[5678] However, it remains unclear which factors, if any, are associated with the early transition from asymptomatic diastolic dysfunction to overt heart failure. The objective of this retrospective, cross-sectional study is to identify echocardiographic markers that may correlate with the development of symptomatic heart failure in the early stages of this disease.

Materials and Methods

Study population

After obtaining approval from our Institutional Review Board, we retrospectively screened all patients hospitalized at our institution between January 2012 and November 2014 for suspected heart failure, who also had a complete transthoracic echocardiogram performed during the index hospitalization demonstrating preserved LV function (n = 425). Of these patients, only 128 patients actually met the American College of Cardiology Foundation/American Heart Association definition for HFpEF, defined as the presence of clinical heart failure, as assessed by physical examination, and a preserved LV ejection fraction (≥50%) on a transthoracic echocardiogram, without the presence of significant valvular abnormalities (moderate or worse by standard echocardiographic guidelines). From this population, all patients with prior valve surgery, history of atrial fibrillation, end-stage renal disease, or significant pericardial disease were excluded from the study.

Eligible patients (n = 40) were then matched in a 1:1 fashion by age and sex to a control population of patients admitted during the same period with a diagnosis of hypertensive urgency, some degree of diastolic dysfunction on a transthoracic echocardiogram performed during the hospitalization, and no evidence of heart failure (n = 40). The medical records of all the eighty patients included in the study were reviewed for prior medical history and clinical data relating to their hospitalization.

Echocardiographic methods

All transthoracic echocardiograms were performed with the patients in the supine position in accordance with the standard guidelines, using commercially available ultrasound systems (General Electric Vivid 7 system, Horten, Norway), and all measurements were performed offline using Echopac PC (GE-Vingmed, Horten, Norway). The complete transthoracic echocardiograms and Doppler echocardiograms for each of the eighty patients included in the study were analyzed in a blinded fashion. Each echocardiographic study used for analysis was obtained during the first 24 h of the patient's heart failure hospitalization. Two echocardiographers extracted all the data for the analysis. A third, level 3 reader reviewed each individual echocardiogram for comparison.

The atrial volumes were obtained using the area-length biplane technique.[9] The contours were methodically traced in the four- and two-chamber views, with care taken to exclude the pulmonary veins and the LA appendage. The maximum and minimum LA areas were measured in ventricular systole just prior to the opening of the mitral valve and in diastole just after closing of the mitral valve. The two shorter LA lengths obtained in the four- and two-chamber views were used in the volume calculation as recommended in the Chamber Quantification Guidelines. The maximal volume and minimal volume indices were computed by dividing the calculated volume by each patient's body surface area. The LA emptying fraction was calculated using the following formula: ([(maximal volume − minimal volume)/maximal volume] ×100).

Pulse-wave Doppler of the peak mitral inflow velocities in early diastole (E) and late diastole (A) was obtained in the apical four-chamber view. Pulse-wave tissue Doppler imaging was also used to obtain the early diastolic (e’) and late diastolic (a’) velocities, at both the septal and lateral sides of mitral annulus. Both measurements were obtained in the apical four-chamber window as recommended by the American Society of Echocardiography.[10] An average e’ was used in the calculation of the E/e’. LA strain analysis by speckle tracking echocardiography was also completed in the two- and four-chamber apical views.

Statistical methods

Data were presented as mean and standard deviation for continuous variables and as frequency and percentage for categorical variables. We utilized Student's independent sample t-test and Pearson's Chi-square analysis to compare means between continuous and categorical variables, respectively. Receiver operating characteristic curves were generated for each echocardiographic variable and utilized to compare their strength of association with heart failure. The c-statistic was calculated, indicating the strength of the discriminatory ability for each echocardiographic variable. Univariate (unadjusted) and multivariate (adjusted) logistic regression analyses were utilized to identify independent associations with heart failure in patients with diastolic dysfunction. Each echocardiographic variable was analyzed with the two best clinical variables (body mass index [BMI] and creatinine) in the adjusted analysis. P < 0.05 was considered statistically significant. In addition, as a supplementary analysis, a generalized propensity score-adjusted analysis was conducted whereby the propensity of levels of the echocardiographic variables was modeled as a function of the baseline characteristics such as BMI, creatinine, age, sex, diabetes, systolic blood pressure, ejection fraction, and beta-blocker use through linear regression, and the resulting probabilities were used as an adjustment factor along with each echocardiographic variable in the final logistic regression model. Goodness of fit of all models was assessed through the Hosmer–Lemeshow test. A Hosmer-Lemeshow P < 0.05 indicates lack of fit. All analyses were performed using SAS Software, version 9.4. Copyright, SAS Institute Inc. SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc., Cary, NC, USA. The inter- and intra-reader variability among the three echocardiographers who participated in this study was evaluated through intraclass correlations and was found to be adequate. P < 0.05 was considered statistically significant with no adjustment for multiple comparisons.

Results

When compared to the hypertensive controls, patients with HFpEF had higher BMI (32.5 ± 11.3 vs. 27.7 ± 5.2, P = 0.019), creatinine (1.79 ± 1.01 vs. 1.02 ± 0.47, P < 0.001), and beta-blocker use (37.5% vs. 23.8%, P = 0.022) [Table 1]. They also had lower systolic (161.4 ± 33.4 vs. 194.1 ± 29.4, P < 0.001) and diastolic (82.8 ± 18.8 vs. 98.6 ± 19.9, P < 0.001) blood pressures [Table 1].

Table 1

Patient characteristics

Variable	Heart failure		P
	Yes (40)	No (40)
Age (years)	72.1±13.0	72.1±12.7	0.993
Women (%)	20 (50)	20 (50)	1.0
BMI (kg/m2)	32.5±11.3	27.7±5.2	0.019
Diabetes mellitus (%)	24 (60)	16 (40)	0.074
Hypertension (%)	38 (96)	39 (98)	0.556
Coronary artery disease (%)	15 (37.5)	13 (32.5)	0.639
Ischemic stroke (%)	3 (7.5)	8 (20)	0.105
Peripheral vascular disease (%)	5 (12.5)	3 (7.5)	0.456
Heart rate	77.9±18.6	76.1±13.8	0.619
Systolic blood pressure	161.4±33.4	194.1±29.4	<0.001
Diastolic blood pressure	82.8±18.8	98.6±19.9	<0.001
Pro-BNP	10,436±21,199	651±632	0.105
Creatinine	1.79±1.01	1.02±0.47	<0.001
Troponin	0.15±0.30	0.06±0.05	0.055
LV ejection fraction (%)	62.9±3.56	64.3±3.31	0.078
Diastolic dysfunction* (%)
None	2.5	0	<0.001
Grade 1	40	95	<0.001
Grade 2	45	2.5	<0.001
Beta-blocker use (%)	30 (75)	19 (47.5)	0.012
Statin use (%)	25 (62.5)	23 (57.5)	0.648
Angiotensin-converting enzyme inhibitor use (%)	23 (57.5)	30 (75)	0.098

*There were insufficient echocardiographic data to make an accurate assessment for the stage of diastolic dysfunction in five patients in the HFpEF group as well as one patient in the control group. BMI: Body mass index, Pro-BNP: Pro-B-type natriuretic peptide, HFpEF: Heart failure with preserved ejection fraction, LV: Left ventricular

As expected, a higher incidence of advanced (Stage II) diastolic dysfunction (45% vs. 2%, P < 0.001) was noted in the heart failure group [Table 1]. In addition, the group of patients that had heart failure was noted to have higher values for right ventricular systolic pressure (42.0 ± 10.4 vs. 31.6 ± 7.1, P < 0.001), LV mass index (126.2 ± 32.6 vs. 110.8 ± 27.8, P = 0.026), E (0.91 ± 0.25 vs. 0.64 ± 0.16, P < 0.001), E/A (1.17 ± 0.62 vs. 0.75 ± 0.29, P < 0.001), and E/e’ (16.4 ± 6.7 vs. 11.3 ± 3.7, P < 0.001) [Table 2]. Similarly, significant differences were observed in all LA size parameters including larger LA maximum volume (71.4 ± 18.6 vs. 45.9 ± 15.8, P < 0.001), maximum volume index (37.8 ± 8.3 vs. 29.0 ± 9.4, P < 0.001), minimum volume (40.2 ± 14.8 vs. 19.8 ± 9.1, P < 0.001), and minimum volume index (22.3 ± 8.5 vs. 13.2 ± 5.8, P < 0.001) in the heart failure group [Table 2]. In addition, LA-emptying fraction (43.6 ± 15.4 vs. 56.4 ± 13.1, P < 0.001) and atrial strain (maximum: 22.3 ± 9.8 vs. 29.3 ± 12.6, P = 0.011, average: 23.1 ± 9.1 vs. 30.4 ± 11.0, P = 0.011), both of which reflect LA function, were noted to be smaller (worse) in the HFpEF group [Table 2].

Table 2

Echocardiographic data

Variable	Heart failure		P
	Yes (40)	No (40)
Right ventricular systolic pressure	42.0±10.4	31.6±7.1	<0.001
LV mass index	126.2±32.6	110.8±27.8	0.026
LA maximum volume (4C)	71.4±18.6	45.9±15.8	<0.001
LA maximum volume index	37.8±8.3	29.0±9.4	<0.001
LA maximum area (4C)	22.5±4.2	17.1±3.7	<0.001
LA maximum length (4C)	5.8±0.7	5.2±0.6	<0.001
LA minimum volume (4C)	40.2±14.8	19.8±9.1	<0.001
LA minimum volume index	22.3±8.5	13.2±5.8	<0.001
LA minimum area (4C)	16.0±3.9	10.2±3.1	<0.001
LA minimum length (4C)	5.2±0.8	4.1±0.7	<0.001
LA-emptying fraction (4C)	43.6±15.4	56.4±13.1	<0.001
Maximum LA strain (4C)	22.3±9.8	29.3±12.6	0.011
Average LA strain (4C)	23.1±9.0	30.4±11.0	0.005
Diastology
E	0.91±0.25	0.64±0.16	<0.001
A	0.89±0.28	0.89±0.21	0.987
E/A	1.17±0.62	0.75±0.29	<0.001
E/e’	16.4±6.7	11.3±3.7	<0.001

LA: Left atrial, LV: Left ventricular, 4C: Four chamber

Of the 54 patients with Grade 1 diastolic dysfunction included in the study, 16 were in the heart failure group (30%) and 38 were in the control group (70%). In this population with early diastolic disease (Grade 1), the left atrial minimum volume index (LAmVI) was also significantly higher in patients who presented with heart failure as compared with hypertensive controls (20.7 ± 8.9 vs. 13 ± 5.7 P = 0.001).

Almost every echocardiographic variable listed in the univariate (unadjusted) analysis in Table 3 had a good correlation with HFpEF with the exception of a’. The parameter that correlated strongly with HFpEF, after adjustment for BMI and creatinine, was LAmVI (c-statistic 0.914) [Table 4], (odds ratio = 1.23 [1.09–1.38], P = 0.001) [Table 5]. All echocardiographic variables demonstrated adequate goodness of fit as indicated by Hosmer–Lemeshow P < 0.05. The results were similar between the BMI/creatinine-adjusted model and the propensity-adjusted model.

Table 3

Univariate (unadjusted) analyses

Effect	n	c-statistic	OR (95% CI)	Effect P
Right ventricular systolic pressure	53	0.815	1.16 (1.06-1.26)	0.001
LV mass index	79	0.651	1.02 (1.00-1.04)	0.026
LA maximum volume (4C)	77	0.845	1.09 (1.05-1.13)	<0.001
LA maximum volume index	72	0.767	1.12 (1.05-1.2)	0.001
LA maximum area (4C)	77	0.829	1.4 (1.2-1.63)	<0.001
LA maximum length (4C)	77	0.788	6.71 (2.54-17.68)	<0.001
LA minimum volume (4C)	77	0.893	1.17 (1.09-1.26)	<0.001
LA minimum volume index	72	0.838	1.26 (1.12-1.42)	<0.001
LA minimum area (4C)	77	0.876	1.61 (1.3-1.98)	<0.001
LA minimum length (4C)	77	0.821	6.3 (2.61-15.22)	<0.001
LA-emptying fraction (4C)	77	0.745	0.94 (0.9-0.97)	0.001
Maximum LA strain (4C)	72	0.698	0.93 (0.88-0.99)	0.018
Average LA strain (4C)	67	0.728	0.92 (0.86-0.98)	0.009
E	77	0.822	559.68 (28.13-11136.62)	<0.001
A	77	0.517	1.02 (0.16-6.38)	0.986
E/A	77	0.772	18.56 (2.98-115.61)	0.002
E/e’	75	0.778	1.29 (1.12-1.49)	<0.001

OR: Odds ratio, CI: Confidence interval, LA: Left atrial, LV: Left ventricular, 4C: Four chamber

Table 4

Multivariate (adjusted) analysis

Effect	Covariate adjusted
	n	c-statistic	OR (95% CI)	Effect P	Hosmer-Lemeshow P
Right ventricular systolic pressure	52	0.881	1.16 (1.05-1.29)	0.004	0.184
LV mass index	78	0.847	1.01 (0.99-1.03)	0.281	0.018
LA maximum volume (4C)	76	0.881	1.08 (1.03-1.12)	0.001	0.426
LA maximum volume index	71	0.883	1.13 (1.04-1.22)	0.002	0.003
LA maximum area (4C)	76	0.885	1.37 (1.14-1.65)	0.001	0.334
LA maximum length (4C)	76	0.860	4.43 (1.56-12.55)	0.005	0.662
LA minimum volume (4C)	76	0.907	1.13 (1.06-1.21)	<0.001	0.404
LA minimum volume index	71	0.914	1.23 (1.09-1.38)	0.001	0.276
LA minimum area (4C)	76	0.903	1.46 (1.19-1.8)	<0.001	0.782
LA minimum length (4C)	76	0.876	4.64 (1.8-11.94)	0.001	0.999
LA-emptying fraction (4C)	76	0.877	0.94 (0.91-0.98)	0.005	0.827
Maximum LA strain (4C)	71	0.887	0.91 (0.84-0.98)	0.011	0.019
Average LA strain (4C)	66	0.894	0.89 (0.82-0.97)	0.007	0.332
E	76	0.884	182.56 (6.93-4809)	0.002	0.185
A	76	0.831	1.21 (0.13-11.37)	0.868	<0.001
E/A	76	0.885	10.01 (1.51-66.25)	0.017	0.319
E/e’	74	0.882	1.27 (1.09-1.49)	0.003	0.631

OR: Odds ratio, CI: Confidence interval, LA: Left atrial, LV: Left ventricular, 4C: Four chamber

Table 5

Final multivariate model

Variable	OR	P
BMI	1.11 (0.99-1.25)	0.074
Creatinine	7.09 (1.43-35.07)	0.016
LA minimum volume index	1.23 (1.09-1.38)	0.001

BMI: Body mass index, OR: Odds ratio, LA: Left atrial

Discussion

The findings of our study are 2-fold. First, in a population of hospitalized patients with varying degrees of diastolic dysfunction and preserved LV systolic function, all the echocardiographic parameters that correlate with increased LV-filling pressures were associated with heart failure hospitalizations. Second, LAmVI had the best association of these echocardiographic parameters.

HFpEF is a complex clinical entity with a magnitude of proposed mechanisms that likely involve interplay between primary structural changes within the heart and related comorbid conditions.[1] Many of these known clinical comorbidities, including renal failure, obesity, and LVH, demonstrated consistent associations with HFpEF in this cross-sectional investigation. However, the most recognized structural mechanism for HFpEF is diastolic dysfunction.

A comprehensive diastology assessment by echocardiography involves multiple parameters, many of which are dependent on loading conditions and sometimes provide conflicting information.[1011] An integral part of this evaluation is the estimation of LV-filling pressures. With worsening diastolic dysfunction, these pressures typically change in a predictable manner, with elevation of LV end-diastolic pressure being the first detectable change, followed by elevations in LA pressure and pulmonary capillary wedge pressure (PCWP).[1012] In patients with normal systolic function, mean LA pressure (and thus PCWP) is estimated by the E/e’ ratio.[12] However, this is a load-dependent variable more reflective of instantaneous filling pressures at the time of the study. In contrast, LA structural changes express chronicity of exposure to elevated filling pressures. This may explain why in prior studies of patients with preserved systolic function, LA dimensions better predicted heart failure hospitalization and other cardiac outcomes.[13]

In this cohort of patients, LA maximum volume index was significantly higher in the group of patients hospitalized for heart failure. This was observed in prior studies in which LA maximum volume >32 ml/m2 was shown to predict both sentinel heart failure hospitalization as well as future hospitalizations.[5614] Similarly, LA maximum volume index >34 ml/m2 has been shown to correlate with other outcomes such as death, atrial fibrillation, and ischemic stroke.[13] For this reason, the new guidelines for the assessment of diastolic dysfunction have added more weight to this variable when grading the disease.[10]

However, LAmVI had the best association with heart failure hospitalization in our patient population [Figure 1]. This novel parameter has been previously described to correlate with PCWP, E/e’, and proBNP better than LA maximum volume index.[7151617] In addition, there is a growing body of evidence suggesting that it may be a better predictor of heart failure and other cardiac end points.[718] In this study, a LAmVI >18 demonstrated the best correlation (94.6% specificity and 61.8% sensitivity) with heart failure hospitalization.

Figure 1

Association between left atrial minimum volume index and heart failure hospitalization in the overall population.

One explanation for this finding is that minimum volume is measured in end-diastole and is thus influenced by LV end-diastolic pressure, known to be one of the earliest changes in diastolic dysfunction.[7] Therefore, changes in LAmVI can be seen in the early stages of diastolic dysfunction, in contrast to LA maximum volumes which do not correlate as well with Grade 1 dysfunction.[1519] The findings from this study support this concept since LAmVI also had the best correlation with heart failure hospitalization in the subgroup of patients with Grade 1 diastolic dysfunction.

In this investigation, LA function (as assessed by LA emptying fraction and atrial strain) showed a modest association with heart failure hospitalization. There is evidence that LA function has some utility in predicting cardiac outcomes.[20] However, the most robust data apply to populations with atrial fibrillation, a subgroup that was excluded from this study.

Limitations

Our study has several limitations inherent to the single-center, retrospective, cross-sectional methodology utilized. Despite finding significant associations between echocardiographic variables and heart failure hospitalization, it is important to note that baseline echocardiograms prior to index hospitalization (development of heart failure) were not available. Thus, we are unable to demonstrate a change over time in the various echocardiographic parameters, and for this reason, we could not make any strong conclusion regarding the variables as predictors of HFpEF. In addition, pulmonary vein flow, an important variable in the diastolic assessment, was not included in our analysis. Finally, a large number of the screened patients were excluded, leaving a fairly small sample size. However, a significant association was demonstrated that should not be overlooked.

Conclusions

HFpEF has become a huge economic burden, and with no proven therapies currently available, the focus should be on monitoring diastolic disease progression in the preclinical phase and prevention of heart failure hospitalization. The LA minimum volume index is a novel echocardiographic parameter that may add value to the diastolic assessment as well as offer prognostic data with regard to predicting heart failure hospitalizations, particularly in the subset of patients without valvular disease or atrial fibrillation. This may potentially identify patients who may benefit from closer surveillance and tighter control of their risk factors. For this reason, these findings should be considered as hypothesis generating, and further studies should be conducted to determine the association between LAmVI and HFpEF.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.