|Home | About | Journals | Submit | Contact Us | Français|
We sought to determine whether left atrial (LA) dysfunction predicts heart failure (HF) hospitalization in subjects with preserved baseline ejection fraction (EF).
Among patients with preserved EF, factors leading to HF are not fully understood. Cross-sectional studies have demonstrated LA dysfunction at the time of HF, but longitudinal data on antecedent atrial function are lacking.
We performed resting transthoracic echocardiography in 855 subjects with coronary heart disease and EF≥50%. Left atrial functional index (LAFI) was calculated as [(LA emptying fraction × left ventricular outflow tract-velocity time integral)/(indexed LA end systolic volume)], where LA emptying fraction was defined as (LA end systolic volume - LA end diastolic volume)/LA end systolic volume. We used Cox models to evaluate the association between LAFI and HF hospitalization.
Over a median follow-up of 7.9 years, 106 participants (12.4%) were hospitalized for HF. Rates of HF hospitalization were inversely proportional to quartile of LAFI: Q1: 47 per 1000 person-years; Q2: 18.3; Q3: 9.6; and Q4: 5.3 (p<0.001). Each standard deviation decrease in LAFI was associated with a 2.6-fold increased hazard of adverse cardiovascular outcomes (unadjusted HR: 2.6, 95% CI 2.1–3.3, p<0.001), and the association persisted even after adjustment for clinical risk factors, NT-proBNP, and a wide range of echocardiographic parameters (adjusted HR: 1.5, 95% CI 1.0–2.1, p=0.05).
LA dysfunction independently predicts HF hospitalization in subjects with coronary heart disease and preserved baseline EF. LAFI may be useful for HF risk stratification, and LA dysfunction may be a potential therapeutic target.
Heart failure (HF) is a major public health problem, affecting 5.8 million people in the United States, with estimated direct and indirect costs of 39 billion dollars in 2010(1). HF is the number one cause of hospitalization in those over the age of sixty-five(2), and approximately half of all HF hospitalizations occur in patients with preserved ejection fractions (HFpEF) (3, 4). Patients with HFpEF have increased mortality and morbidity similar to patients with HF and reduced EF(5, 6). Although medical and device therapies have improved survival for patients with low EF, large randomized trials of traditional HF therapies such as angiotensin blockade have not demonstrated a survival benefit in HFpEF (7–9). The underlying pathophysiology of HFpEF is complex(10), and factors precipitating HF events in patients with preserved EF are not well understood(11). HFpEF has classically been attributed to diastolic dysfunction and left ventricular (LV) stiffness resulting in elevated LV end diastolic pressures(12). However, diastolic dysfunction and LV hypertrophy are also common in patients with hypertension, many of whom never develop clinical HF (13, 14). Therefore, additional discriminating features to identify subjects at highest risk of developing HF are of interest both from a clinical and pathophysiologic standpoint.
Left atrial (LA) remodeling due to overt or subclinical atrial volume or pressure overload could result in decreased atrial systolic function. Atrial dysfunction could lead to impaired atrial emptying, which decreases cardiac output, or it could be an early indicator of cardiac congestion or failure even when EF is preserved. Cross-sectional studies have demonstrated an association between LA dysfunction and HFpEF(14), but longitudinal studies to assess whether LA dysfunction predicts future HF events are lacking. We therefore evaluated the longitudinal association of LA function, as assessed by the left atrial functional index (LAFI), to HF hospitalization in subjects with prevalent coronary heart disease and preserved baseline EF.
The Heart and Soul Study is a prospective cohort study originally designed to investigate psychosocial factors and health outcomes in patients with stable coronary heart disease. Details regarding recruitment methods and study design have been previously published(15). Briefly, between September 2000 and December 2002, we recruited 1,024 outpatients with stable coronary heart disease from two Veterans Administration Medical Centers (Palo Alto and San Francisco), one university medical center (University of California, San Francisco), and nine public health clinics in the Community Health Network of San Francisco. Eligible participants met one or more of the following criteria: (1) history of myocardial infarction; (2) evidence of at least 50% stenosis in 1 or more coronary vessels on cardiac catheterization; (3) evidence of exercise-induced ischemia by treadmill electrocardiogram or nuclear perfusion stress imaging; or (4) a history of coronary revascularization. We excluded those with a history of myocardial infarction in the previous 6 months, inability to walk 1 block, or planning to move out of the local area within 3 years.
Of the 1,024 original study subjects, we excluded the following participants: 110 with baseline EF<50%, 15 with moderate or greater valvular disease, 40 with missing echocardiographic data, and 4 lost to follow-up. The remaining 855 participants are the subjects of this analysis. This study was approved by the institutional review board and all participants provided written, informed consent.
We performed resting transthoracic echocardiography in 855 participants with coronary heart disease and preserved EF (≥50%). These studies were performed in the standard left lateral recumbent and supine positions using an Acuson Sequoia ultrasound system (Siemens Medical Solutions, Mountain View, CA). We obtained standard two-dimensional parasternal short-axis and apical two- and four-chamber views during held inspiration and planimetered these with a computerized digitization system to determine end-diastolic and end-systolic LV volumes by the biplane method of disks. The moments of first mitral valve opening and closing were used to determine end-diastolic and end-systolic LV volumes.
The derivation and validation of LAFI has been previously described(16). LAFI was calculated as [(LA emptying fraction × left ventricular outflow tract-velocity time integral)/(LA end systolic volume index)], where LA emptying fraction was defined as (LA end systolic volume - LA end diastolic volume)/LA end systolic volume (Figure 1). All echocardiograms were performed using a standardized protocol by one of two trained and experienced technicians.
A single experienced reader blinded to clinical information (NBS) interpreted all studies and verified the measurements used for the calculation of LAFI. The reproducibility of LAFI by this reader has been previously described with Bland-Altman analyses, which revealed no significant variation (intra-observer reproducibility: mean difference 0.0059, 95% CI 0.015 to −0.012; inter-observer reproducibility: mean difference 0.0017, 95% CI 0.025 to −0.013)(16).
The primary outcome was time to first HF hospitalization. We conducted annual follow-up interviews with participants or their proxy to inquire about interval hospitalization for “heart trouble”. For any reported event, we retrieved medical records, which two independent and blinded physician adjudicators reviewed. If the adjudicators agreed on the outcome classification, their classification was binding. In the event of a disagreement, a third blinded adjudicator was consulted.
We defined HF as hospitalization for a clinical syndrome based on the Framingham congestive heart failure criteria, which require validation of 2 major or 1 major plus 2 minor criteria. (Major criteria: paroxysmal nocturnal dyspnea, orthopnea, elevated jugular venous pressure, pulmonary rales, third heart sound, cardiomegaly on chest radiograph, pulmonary edema on chest radiograph, weight loss ≥ 4.5 kg in 5 days in response to HF therapy. Minor criteria: peripheral edema, night cough, dyspnea on exertion, hepatomegaly, pleural effusion, heart rate > 120/min)(13, 17).
Age, sex, race, and medical history (including history of HF) were determined by self-reported questionnaire. We measured height and weight, and calculated body mass index (kg/m2). Systolic blood pressure, diastolic blood pressure and heart rate were measured in the supine position after five minutes of rest. We measured serum creatinine, low density lipoprotein, high density lipoprotein and NT-proBNP from fasting blood samples drawn at the baseline study appointment. Estimated glomerular function was calculated by the abbreviated (4-variable) Modification of Diet and Renal Disease Study formula, as follows: estimated GFR = 186 × (serum creatinine−1.154) × (age−0.203) × (0.742 if female) × (1.21 if black)(18). We performed standard 12-lead electrocardiograms on all subjects at the time of enrollment and again after five years of follow-up. Two independent, blinded physicians adjudicated the rhythm of all electrocardiograms. In the event of a disagreement, a third adjudicator was consulted.
From the resting echocardiograms, left atrial volume index (LAVI) was defined as (LA end systolic volume/body surface area). LVEF was calculated as (end-diastolic volume – end systolic volume)/end-diastolic volume(19). LV mass was calculated using the truncated-ellipse method(20) and indexed to body surface area. We defined three categories of diastolic dysfunction based on mitral flow ratios of peak velocities at early rapid filling and late filling at atrial contraction (E/A ratio) and systolic or diastolic dominant pulmonary venous flow: 1) impaired relaxation, defined as an E/A ratio of 0.75 or less and systolic dominant pulmonary venous flow; 2) pseudonormal, defined as an E/A of 0.75 to 1.5 and diastolic dominant pulmonary venous flow; and 3) restrictive, defined as an E/A of 1.5 or greater and diastolic dominant pulmonary venous flow(13). We have previously found differences in rates of cardiovascular outcomes in these three categories of diastolic dysfunction (no diastolic dysfunction, impaired relaxation, pseudonormal/restrictive) (21); therefore, we analyzed diastolic dysfunction as an ordinal variable at these three levels. Because <5% of the study sample had restrictive filling, the pseudonormal and restrictive groups were combined for analysis.
To determine presence of inducible ischemia at baseline, all participants underwent exercise treadmill testing according to a standard Bruce protocol with continuous 12-lead electrocardiogram monitoring. We performed echocardiography immediately before and after exercise. We defined inducible ischemia as the presence of ≥1 new wall motion abnormality at peak exercise.
We estimated pulmonary artery systolic pressure from echocardiography as tricuspid regurgitation gradient plus right atrial pressure. The tricuspid regurgitation jet was visualized with color flow mapping, and the tricuspid regurgitation gradient was measured with continuous wave Doppler. We used the modified Bernoulli equation (delta P=4v2) to calculate gradients from velocities. Right atrial pressure was estimated from the size and respiratory variation of flow in the inferior vena cava.
Participants were divided into quartiles based on their LAFI. We compared differences in baseline characteristics across quartiles using chi-square tests for categorical variables and one-way analysis of variance for continuous variables. Cumulative event-free survival was measured by the method of Kaplan-Meier, and unadjusted differences were compared using the log-rank test. We performed multivariate Cox regression to compare the rate of HF hospitalization across quartiles of LAFI. To determine the independent prognostic value of LAFI, we used incremental multivariate models adjusting for covariates demonstrating an association with LAFI at p≤0.1. We adjusted for age, sex and race (model 1), plus tobacco use, prior revascularization, history of heart failure, atrial fibrillation, estimated glomerular filtration rate, low density lipoprotein cholesterol, angiotensin inhibitors, loop diuretics, and resting heart rate (model 2), plus baseline inducible ischemia (model 3), plus NT-proBNP (model 4), plus diastolic dysfunction, left atrial volume index, left ventricular ejection fraction, left ventricular mass index (model 5). Using the same models, we also examined the rate of HF hospitalization using per SD decrease in LAFI.
Most echocardiographic measures were complete or near complete in all subjects, with the exception of PASP (missing data: left ventricular ejection fraction (N=0), left atrial volume index (N=0), left ventricular mass index (N=6), and diastolic dysfunction (N=19), PASP (N=396)). To determine whether the missing data for PASP was informative, we created a 5-category variable (PASP in quartile 1, 2, 3, 4 or missing) and entered this as an indicator variable. We then tested for interaction to determine whether the association between LAFI and HF differed by age, sex, diabetes, hypertension, obesity, AF, or history of HF, with a cutoff of p≤0.1 considered statistically significant. We also performed a sensitivity analysis in which we adjusted for interim myocardial infarction and AF as time-varying covariates to determine whether the association was independent of interval development of cardiac events. Assessment of assumption of proportional hazards using log-minus-log curves and the Schoenfeld test revealed no violations(22, 23).
We have previously found that LVOT-VTI(24) and LAVI(25), two of the component measures used to derive LAFI, predict HF hospitalization in this cohort. Therefore, we used c-statistics and chi-squared likelihood ratio testing to compare the discrimination of LAFI with each of its individual components. Our group has also found NT-proBNP to be a powerful predictor of HF hospitalization in this cohort(26); therefore, we also compared the discrimination of LAFI with NT-proBNP, which was log-transformed to meet the assumption of linearity. Correlated c-statistics were performed using proportional hazards models with postestimation commands. with All analyses were conducted using STATA (College Station, TX), version 11.0.
During a median follow-up of 7.9 (interquartile range: 4.8–8.1) years, 106 subjects (12.4%) were hospitalized for HF, of whom 71 (67.0%) had no prior history of heart failure. Baseline characteristics of participants across quartiles of LAFI are displayed in Table 1. LAFI was also strongly associated with each of the other prognostic biomarkers and echocardiographic parameters (Table 2).
Event rates increased from 5.3 per 1000 person years in the highest quartile of LAFI to 47.0 per 1000 person years in the lowest quartile (Table 3). Kaplan-Meier survival estimates (Figure 2) revealed early separation of the event-free survival curves (within the first few months), which continued to diverge throughout follow-up. After adjustment for demographics (age, sex, white race), clinical risk factors (tobacco use, prior revascularization, history of heart failure, AF, low density lipoprotein, estimated glomerular filtration rate) medication use (angiotensin inhibitors, loop diuretics), and heart rate, every SD decrease in LAFI increased the adjusted hazard of HF two-fold (HR 2.0, 95% CI, 1.5–2.7; p<0.001). The association also remained independent after further adjustment for log NT-proBNP and a wide range of other echocardiographic measures (per SD decrease in LAFI: HR 1.5, 95% CI, 1.0–2.1; p=0.05) Even after further adjustment for pulmonary artery systolic pressure, point estimates revealed little attenuation (per SD decrease in LAFI: HR 1.4, 95% CI 0.9–2.1, p=0.10) (Table 4). Notably, when PASP was entered into the model as a categorical predictor with missing data treated as a fifth category, data from the missing category was non-informative with respect to the association between LAFI and HF (HR 0.88, 95% CI 0.4–1.9, p=0.88).
To determine whether the association was independent of interval cardiac events, we also performed a sensitivity analysis in which we added interim cardiac events (myocardial infarction and AF) to the adjusted model 2 covariates, and found demonstrated no attenuation of the association (per SD decrease in LAFI: HR 2.2, 95% CI 1.7–2.9; p<0.001).
The association did not vary by age, sex, or the presence of obesity, prior history of HF, or AF (p for interaction > 0.10 for all). We found significant interactions between LAFI and the presence of hypertension (p=0.01) and diabetes (p<0.001). However, stratified analyses revealed the association was present among all subsets, and point estimates were similar in both strata (Hypertension present: n=185, HR 3.0 per SD decrease in LAFI, 95% CI 1.1–8.3 vs. Hypertension absent: n=670, HR 2.0 per SD decrease in LAFI, 95% CI 1.5–2.7; Diabetes present: n=239, HR 1.7 per SD decrease in LAFI, 95% CI 1.2–2.4, Diabetes absent: n=616, HR 2.3 per SD decrease in LAFI, 95% CI 1.6–3.4).
To better characterize the predictive ability of LAFI for incident HF, we also performed a subgroup analysis restricted to subjects with no prior history of HF (n=724) and found results were similar to those of the entire cohort: subjects with LAFI in the lowest quartile had nearly 6 times the rate of incident HF hospitalization compared with those in the highest quartile (adjusted for model 1 covariates: HR 5.8, 95% CI 2.3–14.3; p<0.001), and the rate of HF hospitalization was 80% greater per SD decrease in LAFI (HR 1.8, 95% CI 1.3–2.5, p<0.001). Given the marked preponderance of AF in the lowest quartile of LAFI, we also performed a subgroup analysis limited to only subjects without AF (n=818), which demonstrated no difference compared with the entire cohort (adjusted for model 2 covariates: HR 6.8, 95% CI 3.0–15.0, p<0.001), and the rate of HF hospitalization was 2-fold greater for every 1 SD decrease in LAFI (HR 2.0, 95%CI 1.5–2.7, p<0.001).
The discrimination of LAFI for HF hospitalization was also superior to each of its individual components (unadjusted c-statistics: LAFI 0.73 vs. LVOT-VTI 0.60 (p for comparison < 0.001), LA emptying fraction 0.65 (p<0.001), LAVI 0.69 (p = 0.07)). Cox models also revealed that LAFI provides prognostic value incremental to its component measures (chi-square likelihood ratio testing: p<0.001 for all).
We also compared c-statistics to determine the incremental prognostic value of LAFI when used in conjunction with clinical risk factors and log NT-proBNP. The addition of LAFI to clinical risk factors was significantly more predictive of HF hospitalization than clinical risk factors alone (0.81 for clinical risk factors plus LAFI vs. 0.77 for clinical risk factors alone; p<0.001), or clinical risk factors plus log NT-proBNP (0.85 for clinical risk factors plus log NT-proBNP plus LAFI vs. 0.81 for clinical risk factors plus log NT-proBNP alone; p<0.001).
In a cohort of 855 predominantly male outpatients with stable coronary artery disease and preserved baseline ejection fraction (≥50%), we found that LA dysfunction, as measured by LAFI, is associated with HF hospitalization. This association was independent of age, sex, race, traditional cardiovascular risk factors, heart rate, inducible ischemia, NT-proBNP, and other commonly used echocardiographic parameters (diastolic dysfunction, left atrial volume index, left ventricular ejection fraction, and left ventricular mass index).
LAFI is unique among means of characterizing the LA in that it combines expressions of atrial reservoir function (fractional change), adjusted atrial volume (LAVI), and stroke volume (VTI). For example, an individual with a large LA due to bradycardia will be correctly characterized by LAFI as having normal atrial function because both fractional change and VTI are increased.
Prior studies have demonstrated a correlation between LA volume and diastolic dysfunction(27, 28). LA volume has also been shown to predict incident HF(29). Both rest and reserve LA function have been implicated in HF events in cross-sectional studies of subjects with HFpEF (14, 30). Recent data have also found subjects with HFpEF to have increased atrial contribution to LV filling as a compensatory response to impaired early LV filling during exercise(31). Our demonstration of a longitudinal association between LA function and HF in subjects with preserved baseline EF complements these observations.
Unlike other echocardiographic measures of LA function, LAFI is unique in that it can be measured even in subjects with AF(16). AF is a common comorbidity in subjects with HF; therefore, LAFI is an attractive parameter in this population. As expected, AF was far more common in subjects in the lowest quartile of LAFI compared with those in the highest three quartiles of LAFI. This is consistent with our previous finding that LAFI is low in subjects with AF and increases upon successful cardioversion to sinus rhythm(16). Although LAFI is lower in the setting of AF, we found that the association between LAFI and HF hospitalization was also independent of AF.
Although LA dysfunction is one potential mechanism of HFpEF, several other possible mediators have also been proposed, including inducible ischemia(32), elevated pulmonary artery systolic pressure(33), ventricular-arterial stiffening(34), and higher LV mass index(10, 11, 14). Our findings suggest that the association between LA dysfunction and HF hospitalization in subjects with preserved baseline EF is independent of these factors.
Because the LA contributes up to 30% of stroke volume in healthy individuals, its impairment may precipitate HF. Alternatively, LA dysfunction may simply be a consequence of other pathophysiologic mechanisms which cause HF. If causal, early efforts to prevent or arrest LA dysfunction may be beneficial in those with high risk clinical features. Notably, angiotensin inhibition, which had shown potential to reverse atrial remodeling in animal models(35), did not reduce mortality in clinical trials of subjects with HFpEF(8, 9, 36).
However, other therapies which have shown promise for reversal of atrial remodeling, such as aldosterone antagonists (37) (currently being evaluated for the treatment of HFpEF in the ongoing TOPCAT trial), restoration of sinus rhythm (38, 39), or device therapies (40) could be considered. Regardless of whether improvement of LA function could affect outcomes, LAFI provides prognostic value which is incremental to clinical risk factors and NT-proBNP, and therefore may be useful in risk stratification to identify individuals with preserved baseline EF who are at high risk of HF hospitalization.
Our study has several important limitations. First, our cohort was comprised of predominantly men, which may limit generalizability to women. However, testing for effect modification by sex revealed no difference in the association. Second, although we restricted our analysis to subjects who had a preserved EF at baseline, it is possible that EF or diastolic function may have declined in some subjects prior to HF hospitalization. However, adjustment for the presence of inducible ischemia at baseline and interim cardiac events (myocardial infarction and AF) did not attenuate the association. Third, electrocardiogram data at the time of HF failure presentation was not available; however, adjustment for interim development of AF based on electrocardiograms at year five demonstrated no change in the association. Fourth, the cutpoint of 50% for EF is widely used, but those in the lowest quartile probably include a proportion with established systolic dysfunction. However, further adjustment for EF demonstrated no significant change in the association. Fifth, we evaluated a single measure of resting LA function. Other echocardiographic measures of LA function, including Doppler tissue imaging, segmental atrial function assessment, strain, strain rate, and atrial response to exercise were not examined. Finally, some degree of over-fitting is present in the larger multivariate models; however, results were consistent with the more parsimonious models.
We found that LA dysfunction, as measured by LAFI, is strongly and independently associated with HF hospitalization in patients with preserved baseline EF and stable coronary heart disease. This association remained independent even after adjustment for a wide range of clinical and echocardiographic covariates, and the prognostic value of LAFI was incremental to clinical risk factors and NT-proBNP. LAFI may be useful for HF risk stratification, and LA dysfunction may be a potential therapeutic target.
Sources of Funding
Dr. Welles was supported by a National Research Service Award (1-T32-HP-19025). Dr. Turakhia was supported by a Veterans Health Services Research & Development Career Development Award (CDA09027-1) and an American Heart Association National Scientist Development Grant (09SDG2250647). The Heart and Soul Study was supported by the Department of Veterans Affairs, the National Heart, Lung, and Blood Institute, the American Federation for Aging Research, the Robert Wood Johnson Foundation, and the Ischemia Research and Education Foundation.
Relationships with Industry: None
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.