Author + information
- Received December 28, 2017
- Revision received February 15, 2018
- Accepted February 25, 2018
- Published online April 30, 2018.
- Jin Joo Park, MD, PhDa,
- Jun-Bean Park, MD, PhDb,
- Jae-Hyeong Park, MD, PhDc and
- Goo-Yeong Cho, MD, PhDa,∗ ()
- aCardiovascular Center and Department of Internal Medicine, Seoul National University College of Medicine, Seoul National University Bundang Hospital, Seongnam, Republic of Korea
- bDepartment of Internal Medicine, Seoul National University College of Medicine, Seoul National University Hospital, Seoul, Republic of Korea
- cDepartment of Cardiology in Internal Medicine, Chungnam National University Hospital, Chungnam National University College of Medicine, Daejeon, Republic of Korea
- ↵∗Address for correspondence:
Dr. Goo-Yeong Cho, Cardiovascular Center/Seoul National University Bundang Hospital, Seoul National University, Gumiro 166, Bundang, Seongnam, Gyeonggi-do, Republic of Korea.
Background Heart failure (HF) is currently classified according to left ventricular ejection fraction (LVEF); however, the prognostic value of LVEF is controversial. Myocardial strain is a prognostic factor independently of LVEF.
Objectives The authors sought to evaluate the prognostic value of global longitudinal strain (GLS) in patients with HF.
Methods GLS was measured in 4,172 consecutive patients with acute HF. Patients were categorized as either HF with reduced (LVEF <40%), midrange (LVEF 40% to 49%), or preserved ejection fraction (LVEF ≥50%) and were also classified as having mildly (GLS >12.6%), moderately (8.1% < GLS <12.5%), or severely (GLS ≤8.0%) reduced strain. The primary endpoint was 5-year all-cause mortality.
Results Mean GLS was 10.8%, and mean LVEF was 40%. Overall, 1,740 (40.4%) patients had died at 5 years. Patients with reduced ejection fraction had slightly higher mortality than those with midrange or preserved ejection fraction (41%, 38%, and 39%, respectively; log-rank p = 0.031), whereas patients with reduced strain had significantly higher mortality (severely reduced GLS, 49%; moderately reduced GLS, 38%; mildly reduced GLS, 34%; log-rank p < 0.001). In multivariable analysis, each 1% increase in GLS was associated with a 5% decreased risk for mortality (p < 0.001). Patients with moderate (hazard ratio: 1.31; 95% confidence interval: 1.13 to 1.53) and severe GLS reductions (hazard ratio: 1.61; 95% confidence interval: 1.36 to 1.91) had higher mortality, but LVEF was not associated with mortality.
Conclusions In patients with acute HF, GLS has greater prognostic value than LVEF. Therefore, the authors suggest that GLS should be considered as the standard measurement in all patients with HF. This new concept needs validation in further studies.
Heart failure (HF) is a clinical syndrome with typical symptoms and signs caused by structural and functional cardiac abnormalities (1). Many patients with HF have a left ventricular ejection fraction (LVEF) ≥50% and are categorized as having HF with preserved ejection fraction (HFpEF). Although patients with HFpEF and those with reduced ejection fraction (HFrEF) have a similar prognosis (2), they have distinct underlying etiologies, different degrees of neurohumoral activation, and diverse responses to medical therapy (3,4).
LVEF is one of the most widely used parameters for assessing cardiac function and is a predictor of outcomes; however, the relationship between LVEF and outcomes in patients with HF is inconsistent (5). LVEF represents the percentage change of left ventricular (LV) chamber size, but not myocardial contractility.
Myocardial strain is based on the speckle-tracking method and can be used for the objective and reliable assessment of systolic function (6). It is a simple and feasible method, with excellent reproducibility, and is a strong independent prognostic factor for outcomes in patients with HF, independently of LVEF (7). In this study, we sought to evaluate whether classification of HF according to myocardial strain may better predict clinical outcomes than the clinical phenotype of HF (HFrEF, HF with midrange ejection fraction [HFmrEF], and HFpEF) based on LVEF.
We enrolled 4,312 patients hospitalized for acute HF (AHF) from 3 tertiary university hospitals from January 2009 through December 2016. Patients with signs or symptoms of HF and either lung congestion or objective findings of LV systolic dysfunction or structural heart disease were eligible for the study. We excluded patients who presented with acute coronary syndrome. Echocardiography was performed in 4,237 (98%) patients. We measured both LVEF and global longitudinal strain (GLS). The study protocol was approved by the ethics committee at each hospital. Written informed consent was waived by the institutional review board. The study complied with the Declaration of Helsinki.
All images were obtained using a standard ultrasound machine with a 2.5-MHz probe. Standard techniques were used to obtain M-mode, 2-dimensional, and Doppler measurements in accordance with the American Society of Echocardiography’s guidelines (8). Tissue Doppler–derived peak-systolic, early, and late-diastolic velocities of the septal mitral annulus were recorded. LV end-systolic and end-diastolic volumes were measured from apical 4- and 2-chamber views, and LVEF was calculated by Simpson biplane method.
Echocardiographic images were downloaded from cardiac picture archiving and communication systems in the participating hospitals and archived in digital imaging and communications in medicine format for strain analysis at the strain core laboratory. Of the 4,237 patients with echocardiography, 4,180 (98.6%) had adequate image quality for the strain analysis. Unacceptable images were defined as those with missing views, lacking full cardiac cycles, having >2-segment dropouts, and having significant foreshortening of the LV.
Although recent reports showed that strain analysis software from different vendors had no relevant differences, we used a vendor-independent program. Digitally acquired baseline echocardiographic images in digital imaging and communications in medicine format with acceptable image quality were uploaded to TomTec software (Image Arena 4.6, Munich, Germany) for deformation analyses (2-dimensional cardiac performance analysis), as previously described (9). For deformation analysis, endocardial borders were traced on the end-systolic frame in 3 apical views (4-, 2-, and 3-chamber), with end-systole defined by the QRS complex or as the smallest LV volume during the cardiac cycle. The software tracks speckles along the endocardial border and myocardium throughout the cardiac cycle. Peak longitudinal strain was computed automatically, generating regional data from 6 segments and an average value for each view. For patients in sinus rhythm, analyses were performed on a single cardiac cycle; for patients in atrial fibrillation, strain values were calculated as the average of 3 cardiac cycles. All strain measurements were performed by strain specialists who were blinded to the patients’ other data.
Study variables and definitions
Patients were categorized as having either HFrEF (LVEF <40%), HFmrEF (LVEF 40% to 49%), or HFpEF (LVEF ≥50%) (1). Because GLS is a negative value, we took the absolute value |x| for a simpler interpretation. GLS was divided into tertiles: the ranges for the first (severely reduced strain), second (moderately reduced strain), and third tertiles (mildly reduced strain) were ≤8.0%, 8.1% to 12.5%, and ≥12.6%, respectively. Normal GLS was defined as >20% (10).
The primary outcome was the 5-year all-cause mortality according to GLS and LVEF. The vital statuses of all patients were collected from the national insurance data or national death records.
Assuming an all-cause mortality of 30% (11), a 2-sided log-rank test with an overall sample size of 2,597 subjects (1,299 in the first tertile and 1,298 in the second tertile) achieves 90.0% power at a 0.025 significance level to detect a difference of 6% (12). The proportion of patients lost during follow-up was 0.10. These results are based on the assumption that the hazard rates are proportional. The estimated mortality rate of the group in the first tertile was 0.30. Assuming similar hazard ratios among the 3 groups, a total sample size of 3,897 would be necessary (1,299 patients in each group).
Data were presented as number and frequency for categorical variables and as mean ± SD or median with interquartile range (IQR) for continuous variables. For comparisons among groups, the chi-square test (or Fisher exact test when any expected count was <5 for a 2 × 2 table) was used for categorical variables and the unpaired Student’s t-test, 1-way analysis of variance, Mann-Whitney U test, or Kruskal-Wallis test for continuous variables.
Pearson’s correlation was used to calculate the association between LVEF and GLS. Kaplan-Meier curves were plotted and compared using the log-rank test. A multivariable Cox proportional hazards regression model was used to determine the effect size of LVEF and GLS as independent predictors of all-cause death. We included all variables found to be statistically significant (p < 0.05) in the univariate analysis as covariates in the multivariable analysis, excluding those with >10% of data missing or having multicollinearity with other variables. Variables included in the multivariable analysis were sex, age, systolic blood pressure, New York Heart Association functional class, hypertension, diabetes mellitus, ischemic heart disease, atrial fibrillation, blood urea nitrogen, creatinine, beta-blocker, renin-angiotensin system inhibitor, mineralocorticoreceptor antagonist, GLS, and LVEF. The improvement of reclassification with GLS was assessed using a published method that estimated the net reclassification improvement (13).
A 2-sided p value <0.05 was considered to indicate a statistically significant difference. Statistical tests were performed using SPSS version 22 (IBM, Armonk, New York) and R programming version 3.3.0 (The R Foundation for Statistical Computing, Vienna, Austria). Sample size calculation was performed with PASS 11 (NCSS Statistical Software, Kaysville, Utah).
Baseline characteristics of the study population
Of 4,237 patients with available echocardiography, LVEF could be measured in 4,172 (98.5%) and LV GLS (measurable in all 3 apical views) in 4,180 (98.7%). Regarding the vendor, 60.3%, 21.5%, and 18.1% patients were examined with General Electric (Milwaukee, Wisconsin), Siemens (Mountain View, California), and Phillips (Andover, Massachusetts) devices, respectively. The median time interval between admission and echocardiography was 1 day (IQR: 0 to 2 days).
The mean age was 71 years, 47% were female, and the mean LVEF was 40%. The HF was of ischemic etiology in 33% of the patients and atrial fibrillation was present in 30%. Regarding HF type, 2,195 (51%), 642 (15%), and 1,335 (32%) patients had HFrEF, HFmrEF, and HFpEF, respectively. As for the GLS, 1,404 (33.6%), 1,389 (33.2%), and 1,387 (33.2%) patients were in the first (severely), second (moderately), and third (mildly reduced) GLS tertiles, respectively.
When patients were stratified according to LVEF, those with HFrEF were younger, more likely to be male, had lower blood pressure, and higher natriuretic peptide levels, among other differences (Table 1). Under stratification according to GLS, patients with reduced strain were also more likely to be male, with lower blood pressure, and higher natriuretic peptide levels. In addition, they had lower LVEF.
Correlation between GLS and LVEF
The mean GLS was 10.8 ± 5.0%, and the mean LVEF was 40 ± 15%. There were 132 patients (5%) with normal GLS, defined as >20%; in contrast, 1,022 (24.5%) patients had a normal ejection fraction (EF), defined as ≥55%. Although there was a moderate but highly significant correlation between LVEF and GLS (r = 0.69; p < 0.001), GLS was distributed widely at a given LVEF level (Figure 1A). The proportion of patients with HFrEF decreased as GLS increased (i.e., 54%, 37%, and 38% in the first, second, and third GLS tertiles, respectively) (Figure 1B).
Both B-type natriuretic peptide (BNP) and N-terminal pro–BNP levels increased as GLS decreased (BNP: 1,847 ± 1,734 pg/ml, 1,943 ± 3,873 pg/ml, and 1,044 ± 1,428 pg/ml; N-terminal pro–BNP: 11,743 ± 12,181 pg/ml, 9,350 ± 11,044 pg/ml, and 6,234 ± 10,494 pg/ml for the first, second, and third GLS tertiles, respectively).
The median follow-up duration was 31.7 months (IQR: 11.6 to 54.4 months). Overall, 1,740 (40.4%) patients had died at 5 years. They had more unfavorable baseline characteristics, such as old age and comorbidities. There was no difference in LVEF between the patients who were still living and those who had died, whereas GLS was lower in the patients who had died (Table 2). In Kaplan-Meier survival analysis, patients with HFrEF had higher mortality (p = 0.031) and a higher composite of all-cause mortality and hospitalization for HF (Figure 2). Regarding GLS, patients with severely reduced GLS had the worst clinical outcomes (Figure 3). Under stratification according to LVEF, mortality increased with the severity of GLS reduction in all patients, and in the HFrEF, HFmrEF, and HFpEF subgroups, indicating that strain has a prognostic value that is independent of LVEF. In restricted cubic splines, the mortality decreased with decreased strain, although its relationship with LVEF was not prominent (Central Illustration).
In multivariable analysis, each 1% absolute increase in GLS was associated with a 5% decreased risk for mortality (hazard ratio [HR]: 0.95; 95% confidence interval [CI]: 0.93 to 0.96; p < 0.001) (Online Table 1). Compared with patients with mildly reduced strain (third tertile), patients with moderately (second tertile) (HR: 1.31; 95% CI: 1.13 to 1.53) and severely (first tertile) (HR: 1.61; 95% CI: 1.36 to 1.91) reduced strain had a higher risk for 5-year all-cause death. In contrast, LVEF was not associated with mortality after adjustment (Table 3). Similar results were obtained when the patients were stratified according to LVEF tertiles (Online Figure 1, Online Table 2).
We performed 3 additional analyses to examine the superiority of prognostic information of GLS over LVEF. GLS had a higher area under the curve to determine the 5-year all-cause mortality than LVEF (C-index, 0.59 vs. 0.51; p < 0.001). Patient classification was significantly improved when GLS was added to LVEF (continuous net reclassification improvement: 12%; 95% CI: 6.3% to 15.7%; p < 0.001; integrated discrimination improvement: 1.9%; 95% CI: 1.1% to 2.8%; p < 0.001).
We performed systematic measurement of GLS in 4,172 consecutive patients with HFrEF, HFmrEF, and HFpEF for the first time. To the best of our knowledge, our study is the largest and the only confirmative study on the prognostic value of GLS in patients with AHF across all HF types. We found a moderate, but highly significant correlation between GLS and LVEF; however, for a given LVEF there was a wide distribution of GLS. More than 95% of patients with HF, including those with HFpEF, had reduced GLS. Nonetheless, patients with HFrEF had lower and those with HFpEF had higher GLS. GLS, but not LVEF, was an independent predictor of mortality in patients with HFrEF, HFmrEF, and HFpEF. Hence, GLS should be considered as the standard measurement for the prediction of mortality in patient with AHF. This new concept needs extensive validation in further clinical trials and practice, but it may also explain some of the paradoxical phenomena observed in HF management.
HFpEF is not HF with preserved systolic function
The most widely used and accepted single marker for evaluating cardiac function is LVEF measured during echocardiography; it enables visualization and quantification of LV function. However, there are patients with the clinical syndrome of HF, but with normal or near normal EF, who are classified as having HFpEF (1). Despite their preserved EF, their prognosis is similar to that of patients with reduced EF (2,4).
In earlier guidelines, HFpEF has been described as diastolic HF, emphasizing the impaired LV diastolic function responsible for HF manifestation. However, in a CHARM echocardiographic substudy (CHARMES), diastolic dysfunction was observed in only 67% of patients, and only moderate-to-severe diastolic dysfunction (44% of the study population) was an independent predictor of adverse outcomes (14). Furthermore, in a community-based study, 28% of the randomly selected residents had diastolic dysfunction, whereas only 2.2% had HF (15). Therefore, it is of clinical interest to identify who among those with diastolic dysfunction develops HF.
Systolic and diastolic dysfunction are interrelated. The underlying pathophysiology is an aberrant calcium homeostasis (i.e., impaired release and uptake of intracellular calcium) (16). Thus, both systolic and diastolic dysfunction are expected to be present in HFpEF and HFrEF. One of our key findings is that 84% of the patients with HFpEF had decreased GLS (<20%) and the degree of decreased GLS was related to the prognosis, implying that a significant proportion of patients with HFpEF have reduced systolic function. In an echocardiographic substudy of the TOPCAT (Treatment of Preserved Cardiac Function Heart Failure with an Aldosterone Antagonist) trial, patients with HFpEF had lower longitudinal strain compared with age-, sex-, and race-matched elderly individuals without HF (17).
Outcomes according to GLS and LVEF
Our principal findings are related to mortality. There was a very small difference in mortality among patients with HFrEF (41%), HFmrEF (38%), and HFpEF (39%), whereas morality was higher in patients in the first GLS tertile (49%) than in those in the second (38%) or third GLS tertile (34%). It is noteworthy that LVEF itself had a very limited prognostic value; in contrast, the GLS tertiles showed excellent prognostic value.
Because LVEF and GLS correlated significantly, we performed stratification to examine the sole effect of one parameter independently of the other. When stratified according to GLS (data not shown), LVEF did not have any prognostic value. Strikingly, however, when patients were stratified according to LVEF, their mortality increased from mildly to severely reduced strain. Furthermore, every 1% decrease in GLS was associated with a 5% increased risk for death, after adjustment for significant covariates.
LVEF is derived from geometric deformation of the endocardium (18). For its accurate measurement, appropriate echocardiographic views with acceptable quality are necessary. In a core laboratory analysis of baseline echocardiographic studies in the STICH trial, which included patients with ischemic cardiomyopathy and LVEF <35%, Oh et al. (19) showed that 18% of the patients had an EF ≥35%. Furthermore, in 73% of the patients, LVEF was measured using the Simpson method because of poor imaging quality, suggesting that LVEF has poor reproducibility and feasibility (19). LVEF also has the intrinsic limitation that it does not necessarily represent myocardial contractility. Even in cases of definite regional wall motion abnormalities caused by myocardial infarction, LVEF can be within the normal range because of compensatory mechanisms; however, those patients obviously have increased neurohumoral activation, which has a deleterious effect on clinical outcomes (20). This may explain the inconsistent and controversial relation between LVEF and clinical outcomes. Furthermore, LVEF is dependent on geometry and the expertise of operators, among other factors. Strain, in contrast, measures the myocardium directly. We showed that, using digital imaging and communications in medicine format files, regardless of vendor, GLS could be successfully measured in >98% of patients at the core laboratory. Strain is an objective method with high feasibility.
Generally, the precise classification of disease helps to diagnose, initiate appropriate treatment, and predict patients’ outcomes. Considering that establishing a diagnosis of HFpEF with clinical criteria is difficult, most patients with HFpEF had reduced GLS, and the prognostic value of LVEF is inconclusive, classifying HF by GLS may be a novel approach that could overcome the limitations of LVEF.
Study strengths and limitations
There are several small-sized studies that investigated the prognostic value of GLS versus LVEF (21,22). However, those studies included only patients with HFrEF (7,23) or HFpEF (17,24,25). To the best of our knowledge, this is the first and largest study to compare outcomes according to GLS and LVEF across the entire spectrum of patients with HF (i.e., HFpEF, HFmrEF, and HFpEF). In an echocardiographic substudy of the PARAMOUNT trial, Shah et al. (17) showed that patients with HFpEF had decreased longitudinal strain. However, because of its small sample size, the study merely showed a correlation between longitudinal strain and N-terminal pro–BNP levels. By contrast, our study was adequately powered to evaluate hard clinical outcomes.
Nevertheless, there are several limitations. First, the assessment of LVEF could not be standardized; therefore, there may have been variations among different operators and different techniques, leading to misclassification of HFrEF, HFmrEF, and HFpEF. Measurement of LVEF with 3-dimensional echocardiography may have overcome this limitation (26).
Second, we do not know whether HF classification according to GLS can better guide therapy. Trials with disease-modifying drugs have shown neutral results in patients with HFpEF (27). Further investigation to identify patient subgroups that may respond to medical therapy is required. Third, because we enrolled only patients admitted for AHF, it is unknown whether the current study findings can be extrapolated to patients with chronic stable HF. Because of the wide inclusion criteria, the study population was heterogeneous, and many of the patients may have multiple comorbidities. Finally, although myocardial strain and strain rate are currently excellent markers for echocardiographic myocardial function, it has not yet been accepted as the gold standard method.
Almost all patients with HF had reduced strain, regardless of their LVEF. GLS has greater prognostic value than LVEF. Therefore, we suggest GLS should be considered as the standard measurement in all patients with HF. This new concept needs extensive validation in further clinical trials before being accepted in clinical practice. In addition, we do not know whether HF classification according to GLS can better guide therapy. Trials with disease-modifying drugs have shown neutral results in patients with HFpEF.
COMPETENCY IN PATIENT CARE AND PROCEDURAL SKILLS: In patients with AHF, GLS has greater prognostic value as a predictor of mortality than LVEF.
TRANSLATIONAL OUTLOOK: Further clinical studies are necessary to validate the prognostic value of GLS in patients with acute and chronic HF and to evaluate how classification of HF severity by GLS can be used to guide therapy.
The authors thank Sun-Hwa Kim, PhD, for her assistance with the statistical analyses.
All authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- acute heart failure
- B-type natriuretic peptide
- confidence interval
- ejection fraction
- global longitudinal strain
- heart failure
- heart failure with midrange ejection fraction
- heart failure with preserved ejection fraction
- heart failure with reduced ejection fraction
- left ventricular
- left ventricular ejection fraction
- Received December 28, 2017.
- Revision received February 15, 2018.
- Accepted February 25, 2018.
- 2018 American College of Cardiology Foundation
- Ponikowski P.,
- Voors A.A.,
- Anker S.D.,
- et al.
- Butler J.,
- Fonarow G.C.,
- Zile M.R.,
- et al.
- Kang S.H.,
- Park J.J.,
- Choi D.J.,
- et al.,
- for the KorHF Registry
- Curtis J.P.,
- Sokol S.I.,
- Wang Y.,
- et al.
- Cho G.Y.,
- Marwick T.H.,
- Kim H.S.,
- Kim M.K.,
- Hong K.S.,
- Oh D.J.
- Kraigher-Krainer E.,
- Shah A.M.,
- Gupta D.K.,
- et al.,
- for the PARAMOUNT Investigators
- Motoki H.,
- Borowski A.G.,
- Shrestha K.,
- et al.
- Sengeløv M.,
- Jørgensen P.G.,
- Jensen J.S.,
- et al.
- Persson H.,
- Lonn E.,
- Edner M.,
- et al.,
- for the Investigators of the CHARM Echocardiographic Substudy-CHARMES
- Kranias E.G.,
- Hajjar R.J.
- Shah A.M.,
- Claggett B.,
- Sweitzer N.K.,
- et al.
- Stokke T.M.,
- Hasselberg N.E.,
- Smedsrud M.K.,
- et al.
- Kalam K.,
- Otahal P.,
- Marwick T.H.
- Morris D.A.,
- Ma X.X.,
- Belyavskiy E.,
- et al.
- Donal E.,
- Lund L.H.,
- Oger E.,
- et al.,
- for the KaRen investigators
- Freed B.H.,
- Daruwalla V.,
- Cheng J.Y.,
- et al.
- Thavendiranathan P.,
- Grant A.D.,
- Negishi T.,
- Plana J.C.,
- Popović Z.B.,
- Marwick T.H.