Left Ventricular Ejection Time is an Independent Predictor of Incident Heart Failure in a Community based Cohort

Abstract

Aims:

Systolic time intervals change in the progress of cardiac dysfunction. The usefulness of left ventricular ejection time (LVET) to predict cardiovascular morbidity, however, is unknown.

Methods and Results:

We studied middle-aged African-Americans from one of four cohorts of the Atherosclerosis Risk in Communities study (Jackson cohort, n=1,980) who underwent echocardiography between 1993 and 1995. LVET was measured by pulsed-wave Doppler of the left ventricular outflow tract and related to outcomes.

A shorter LVET was associated with younger age, male sex, higher diastolic blood pressure (BP), higher proportion of diabetes, higher heart rate, higher blood glucose levels and worse fractional shortening (FS). During a median follow-up of 17.6 years, 384 (19%) had incident heart failure (HF), 158 (8%) had a myocardial infarction (MI), and 587 (30%) died. In univariable analysis, a lower LVET was significantly associated with increased risk of all events (p<0.05 for all). However, after multivariable adjustment for age, sex, hypertension, diabetes, body mass index, heart rate, systolic and diastolic BP, FS and left atrium diameter, LVET remained an independent predictor only of incident HF (HR1.07 (1.02–1.14), P=0.010, per 10ms decrease). In addition, LVET provided incremental prognostic information to the known risk factors included in the Framingham risk score, in regard to predicting all outcomes except for MI.

Conclusion:

LVET is an independent predictor of incident HF in a community-based cohort and provides incremental prognostic information on the risk of future HF and death when added to known risk prediction models.

Introduction

The left ventricular ejection time (LVET) is defined by the opening and closing of the aortic valve, which in turn mainly is determined by the pressure differences across the valve. In the ailing heart, the LVET will change during disease progression as previously demonstrated in patients with ischemic heart disease, heart failure, hypertension and aortic stenosis^1–3. As left ventricular (LV) function deteriorates, the ability of the heart to produce contractile force is attenuated and the rate of the LV pressure rise (LV dP/dt) during the isovolumic contraction decreases, resulting in a prolongation of isovolumic contraction time (IVCT)^4–6. Furthermore, the ability of the ailing heart to maintain a high LV pressure during the ejection period decreases, resulting in reduction in the LVET^4–6. Additionally, the LVET will also shorten with LV deterioration simply as the result of the prolonged IVCT which induces a delayed onset of ejection⁶.

Systolic time intervals, including LVET, were used frequently in the past because they were easily obtained by phonocardiography^6–8, and are also easily obtained with Doppler echocardiography⁹. The LVET has been demonstrated to identify impaired cardiac function in patients with ischemic heart disease¹⁰, hypertension¹¹, primary pulmonary hypertension^12,13 and heart failure (HF)^7,14,15. In addition, the LVET has previously been demonstrated to be a strong predictor of cardiovascular outcome in selected patient populations, especially in patients with ischemic heart disease^8,10,16–19 and primary pulmonary hypertension^12,13, but whether this measure has prognostic utility in a general population free of cardiovascular disease is unknown. In addition, it is not well understood which cardiovascular outcomes a shorter LVET is associated with.

Methods

Study Population:

The study population consists of the Jackson cohort of the Atherosclerosis Risk in Communities (ARIC) study. The ARIC study is an ongoing, prospective observational study of the natural history of atherosclerotic diseases and cardiovascular risk factors. Detailed study rationale, design, and procedures have been previously published²⁰. The original cohort was recruited between 1987–1989 using probability sampling of middle aged (45–64 years old) men and women from 4 communities in the United States (Forsyth County, NC; Jackson, MS; Minneapolis, MN; and Washington County, MD). The Jackson field center enrolled an entirely African-American cohort, the Jackson cohort. Subsequent follow up visits occurred at approximately 3 year intervals up to Visit 4 (1996–1999), with annual telephone interviews conducted between visits. Institutional review boards approved the study and informed consent was obtained from all participants.

Transthoracic echocardiography was performed only in the Jackson cohort during visit 3 (1993–1996). Of 2,445 participants who underwent transthoracic echocardiography, 2,257 had LVET measured. After excluding participants with previous HF, missing HF data, and known coronary heart disease (n=277), our study population included 1980 participants.

Demographics:

Hypertension, diabetes mellitus, coronary heart disease, stroke, smoking status, and medication use was defined as previously described in ARIC²¹. Estimated glomerular filtration rate, hematologic parameters, lipids, and glucose were measured according to standardized protocols.

Echocardiography:

Echocardiograms were recorded by four trained sonographers and interpreted by experienced cardiologists in the Echocardiography Reading Center located at the University of Mississippi Medical Center. Two-dimensional, M-mode, and Doppler images were acquired with an Acuson 128XP/10c cardiac ultrasound machine with 2.5, 3.5, and 5.0 MHz transducers (Acuson, Malvern, PA). Measurements were performed by the interpreting physician who was blinded to the participants’ clinical data. LV end diastolic diameter (LVEDD), LV end systolic diameter, septal and posterior wall thickness, and left atrial dimension (LAD) were measured from 2-dimensional images according to American Society of Echocardiography criteria²². LVET was measured as the duration of the flow through the left ventricular outflow tract (LVOT) as assessed by pulsed-wave Doppler. The LVET corrected by the RR-interval obtained from the same cardiac cycle was calculated as LVET/RR.

Outcomes:

The follow up period was defined as the time elapsed from the date of echocardiography to the date of event, date of last contact for those lost to follow-up, or end of 2012. Incident HF was defined as the occurrence of a hospitalization with an ICD-9 discharge code 428 in any position or a death certificate with either an ICD-9 code 428 in any position or an ICD-10 code 150 in any position. Incident MI was defined as definite or probable hospitalization for MI based on committee adjudication of abstracted hospitalization records including chest pain symptoms, electrocardiograms (ECGs), and cardiac enzymes. Death was ascertained by annual phone call follow-up or through health department death certificate files²³. The composite outcome of incident HF, incident MI or death was also assessed.

Statistical analysis:

Baseline characteristics across LVET quartiles were compared with trend tests (for continuous Gaussian distributed variables obtained by regression analysis, by an extension of the Wilcoxon rank-sum test²⁴ for continuous non-Gaussian distributed variables and by a chi-square test for proportions (Cochran-Armitage trend test)). Rates of all events were calculated (number of events divided by person-time at risk) and stratified by quartiles of LVET. Hazard ratios (HR) were calculated by Cox proportional hazards regression analyses using time since visit 3. The association between LVET, LVET/RR and outcomes were assessed using the variables as linear predictors. In addition, departure from linearity for the association between LVET and outcomes were tested using restricted cubic splines with the number of knots selected according to the value associated with the lowest Akaike information criterion (AIC) value. The association was further assessed using a piecewise linear model, with separate linear relationships considered for SET values < 350ms and for SET values ≥ 350ms. The predictive capabilities of LVET was assessed in a univariable model, an age and gender adjusted model, a model including the important demographic determinants of LVET (model 2a), a model including the variables included in the Framingham risk score (2b) and a model including the important demographic determinants of LVET and echocardiographic measures of systolic and diastolic function. Harrell’s c-statistics²⁵ obtained from Cox proportional hazards regression models including the variables from the Framingham Risk score²⁶ (age, gender, total cholesterol, HDL cholesterol, smoking status and systolic blood pressure) and the SCORE risk chart²⁷ (age, gender, cholesterol, smoking status and systolic blood pressure) were calculated and compared with the c-statistics obtained from Cox proportional hazards models also including the LVET in order to test the incremental prognostic performance of the models when adding LVET to the parameters from the conventional risk scores. C-statistics were compared using a transformation of the equivalent Somers’ D parameters²⁸. Similarly, continuous and categorical net reclassification improvement (NRI) and integrated discrimination improvement (IDI) were obtained when adding LVET to the parameters included in Framingham Risk score using Cox models and 15 years of follow-up (Supplemental Table 1). Colinearity and interaction was assessed for LVET and all covariates included in the final multivariable model (Model 3). A p-value ≤ 0.05 in 2-sided test was considered statistically significant. No corrections for multiple testing were performed. All analyses were performed with STATA Statistics/Data analysis, SE 12.0 (StataCorp, Texas,USA).

Results

The population demographics stratified by quartiles of LVET are shown in Table 1. Briefly, the participants had a mean age of 59 ± 6 years, 36% (n=718) were males, 57% (n=1128) had hypertension, 22% (n=718) had diabetes, the mean systolic and diastolic BP were 140 ± 20 mmHg and 82 ± 10 mmHg, respectively, the mean heart rate was 66 ± 10 beats per minute and the mean LVET was 342 ± 35 ms. Participants with a shorter LVET were younger, more likely male, had higher diastolic blood pressure, a higher proportion of diabetes, a higher heart rate, higher blood glucose levels, lower fractional shortening (FS), and a larger LV end-systolic dimension (Table 1).

Table 1.

Clinical and echocardiographic parameters by quartiles of left ventricular ejection time

	1st Quartile	2nd Quartile	3rd Quartile	4th Quartile	P-value
LVET (ms)	<322	322–340	341–360	>360	for
	n=495	n=497	n=506	n=482	trend
Age	57.9 ± 5.6	58.3 ± 5.7	58.7 ± 5.6	59.6 ± 5.7	<0.001
Male gender	246 (50%)	186 (37%)	167 (33%)	119 (25%)	<0.001
Systolic Blood Pressure	140 ± 21	139 ± 19	139 ± 20	142 ± 21	0.10
Diastolic Blood Pressure	85 ± 11	83 ± 10	82 ± 10	81 ± 10	<0.001
Hypertension	313 (63%)	263 (53%)	268 (54%)	284 (59%)	0.21
Diabetes	143 (29%)	92 (19%)	95 (19%)	98 (21%)	0.002
BMI	30 ± 6	30 ± 6	30 ± 6	31 ± 6	0.33
Heart Rate	72 ± 11	66 ± 9	64 ± 9	60 ± 9	<0.001
Smoking status
- Current	120 (24%)	102 (21%)	87 (17%)	82 (17%)	<0.001
- Former	181 (37%)	148 (30%)	151 (30%)	135 (28%)
- Never	191 (39%)	244 (49%)	262 (52%)	264 (55%)
QRS interval (ms)	95 ± 15	95 ± 14	95 ± 14	96 ± 16	0.49
Total Cholesterol (mg/dL)	207 ± 40	207 ± 39	208 ± 39	207 ± 39	0.80
Total Glucose (mg/dL)	105 (95–133)	101 (93–113)	102 (94–114)	99 (92–113)	<0.001
Echocardiographic Measures:
Fractional Shortening (FS) (%)	32.6 ± 9.4	34.9 ± 8.7	35.0 ± 8.4	36.0 ± 8.0	<0.001
LV end diastolic diameter (cm)	4.7 ± 0.6	4.6 ± 0.6	4.6 ± 0.6	4.7 ± 0.6	0.79
LV end systolic diameter (cm)	3.0 ± 0.7	2.8 ± 0.6	2.8 ± 0.5	2.8 ± 0.5	<0.001
Left Atrium Diameter (LAD) (cm)	3.4 ± 0.6	3.3 ± 0.5	3.3 ± 0.5	3.4 ± 0.5	0.22

LVET – Systolic Ejection Time; BMI – Body Mass Index; LV – Left Ventricle.

Systolic ejection time and outcome

During a median follow-up of 17.6 years, outcomes included 384 with incident HF (19%), 158 with MI (8%), and 587 were deceased (30%). Participants in the lowest LVET quartile had a 55% increased risk of the composite endpoint (HF, MI or death) compared to participants in the highest quartile (1^st quartile vs. 4^th quartile HR 1.55 (1.28–1.88), p<0.001)(Figure 1, Table 2 and 3). In comparison, the risk of incident HF was 66% higher in the 1^st quartile (1^st quartile vs. 4^th quartile HR 1.66 (1.26–2.19), p<0.001) and risk of mortality 65% higher in the 1^st quartile (1^st quartile vs. 4^th quartile HR 1.65 (1.31–2.07), p<0.001)(Figure 2 and Table 2) compared to participants in the highest quartile. The risk of MI did not differ significantly between participants in the lowest compared to the highest quartile (1^st quartile vs. 4^th quartile HR 1.46 (0.94–2.26), p=0.09)(Figure 2 and Table 2).

Figure 1: Association between Left Ventricular Ejection Time and risk of the Composite endpoint (HF, MI or death).

Depicting the unadjusted incidence rate of the composite endpoint (events per 1000 person years) on the left y-axis and systolic ejection time (LVET) on the x-axis. The percentage of the population corresponding to the histogram is displayed on the right y-axis. The black curve depicts the incidence, with 95% confidence intervals of the estimates (P for overall trend <0.001 and P for non-linearity <0.001). Also depicted is the histogram of the LVET in our population (n=1980).

LVET – Left Ventricular Ejection Time; HF – Heart Failure; MI – Myocardial Infarction.

Table 2.

Event rates by quartiles of left ventricular ejection time

	1st Quartile	2nd Quartile	3rd Quartile	4th Quartile	P-value for
Left Ventricular Ejection Time (ms)	<322	322–340	341–360	>360	trend
	n=495	n=497	n=506	n=482
Event rate of composite endpoint (per 1000 person years) (95% CI)	35 (31; 39)	23 (20; 26)	24 (21; 28)	23 (20; 27)	<0.001
Event rate of heart failure (per 1000 person years) (95% CI)	17 (14; 21)	11 (9; 14)	12 (9; 14)	11 (9; 13)	0.001
Event rate of myocardial infarction failure (per 1000 person years) (95% CI)	6 (5; 9)	4 (3; 6)	6 (4; 7)	4 (3; 6)	0.25
Event rate of mortality (per 1000 person years) (95% CI)	25 (21; 28)	17 (14; 20)	12 (99; 14)	11 (9; 13)	<0.001

Table 3.

Left ventricular ejection time and the association with outcomes

	Combined endpoint (782 events) Hazard Ratio (95% CI)	P-value	HF (384 events) Hazard Ratio (95% CI)	P-value	HF/Death (733 events) Hazard Ratio (95% CI)	P-value	MI* (158 events) Hazard Ratio (95% CI)	P-value	MI/Death* (664 events) Hazard Ratio (95% CI)	P-value	Death (586 events) Hazard Ratio (95% CI)	P-value
Unadjusted model (n=1980):
LVET per 10 ms decrease							1.06	0.024	1.05	<0.001
							(1.01–1.11)		(1.02–1.07)
							C-stat 0.54		C-stat 0.55
LVET per 10 ms decrease below 350 ms	1.11	<0.001	1.12	<0.001	1.11	<0.001					1.11	<0.001
	(1.07–1.14)		(1.07–1.17)		(1.07–1.14)						(1.07–1.14)
	P<0.001		P<0.001		P<0.001						P<0.001
LVET per 10ms increase above 350ms	1.02		1.02		1.02						1.02
	(0.99–1.05)		(0.97–1.06)		(0.99–1.05)						(0.99–1.06)
	P=0.30		P=0.49		P=0.21						P=0.19
	C-stat 0.55		C-stat 0.56		C-stat 0.55						C-stat 0.56
LVET/RR per 10 % increase	1.24	0.003	1.38	0.001	1.32	<0.001	0.93	0.67	1.21	0.015	1.29	0.002
	(1.08–1.42)		(1.13–1.67)		(1.15–1.53)		(0.68–1.28)		(1.04–1.40)		(1.10–1.51)
	C-stat 0.54		C-stat 0.56		C-stat 0.55		C-stat 0.50		C-stat 0.53		C-stat 0.54
Multivariable Model 1 (n=1980):
LVET per 10 ms decrease							1.06	0.033	1.05	<0.001
							(1.00–1.11)		(1.02–1.07)
							C-stat 0.61		C-stat 0.65
LVET per 10 ms decrease below 350 ms	1.10	<0.001	1.13	<0.001	1.11	<0.001					1.11	<0.001
	(1.07–1.14)		(1.09–1.18)		(1.08–1.15)						(1.07–1.14)
	P<0.001		P<0.001		P<0.001						P<0.001
LVET per 10 ms increase above 350 ms	1.01		1.00		1.01						1.02
	(0.98–1.04)		(0.95–1.05)		(0.98–1.05)						(0.98–1.06)
	P=0.59		P=0.92		P=0.50						P=0.36
	C-stat 0.65		C-stat 0.66		C-stat 0.66						C-stat 0.67
LVET/RR per 10% increase	1.40	<0.001	1.42	0.001	1.48	<0.001	1.05	0.59	1.41	<0.001	1.49	<0.001
	(1.22–1.62)		(1.16–1.74)		(1.28–1.72)		(0.76–1.45)		(1.20–1.64)		(1.27–1.76)
	C-stat 0.64		C-stat 0.64		C-stat 0.65		C-stat 0.59		C-stat 0.65		C-stat 0.66
Multivariable Model 2a (n=1887):
LVET per 10 ms decrease							1.09	0.007	1.04	0.012
							(1.02–1.16)		(1.01–1.07)
							C-stat 0.71		C-stat 0.69
LVET per 10 ms decrease below 350 ms	1.08	<0.001	1.10	<0.001	1.08	<0.001					1.07	0.006
	(1.04–1.12)		(1.04–1.15)		(1.04–1.12)						(1.03–1.12)
	P<0.001		P=0.001		P<0.001						P=0.001
LVET per 10 ms increase above 350 ms	1.00		1.00		1.01						1.02
	(0.96–1.04)		(0.95–1.07)		(0.97–1.05)						(0.97–1.06)
	P=0.89		P=0.89		P=0.71						P=0.50
	C-stat 0.70		C-stat 0.74		C-stat 0.71						C-stat 0.71
LVET/RR per 10% increase	1.06	0.54	0.85	0.23	1.08	0.43	0.86	0.49	1.13	0.21	1.15	0.20
	(0.88–1.27)		(0.64–1.11)		(0.89–1.31)		(0.57–1.32)		(0.93–1.38)		(0.93–1.41)
	C-stat 0.69		C-stat 0.74		C-stat 0.71		C-stat 0.70		C-stat 0.69		C-stat 0.70
Multivariable Model 2b (n=1931):
LVET per 10 ms decrease							1.04	0.12	1.05	<0.001
							(0.99–1.09)		(1.02–1.07)
							C-stat 0.71		C-stat 0.70
LVET per 10 ms decrease below 350 ms	1.08	<0.001	1.11	<0.001	1.09	<0.001					1.08	<0.001
	(1.05–1.12)		(1.06–1.16)		(1.06–1.13)						(1.05–1.12)
	P<0.001		P<0.001		P<0.001						P<0.001
LVET per 10 ms increase above 350 ms	0.99		0.99		0.99						1.00
	(0.95–1.03)		(0.93–1.05)		(0.95–1.03)						(0.95–1.04)
	P=0.68		P=0.72		P=0.72						P=0.96
	C-stat 0.70		C-stat 0.72		C-stat 0.70						C-stat 0.71
LVET/RR per 10% increase	1.38	<0.001	1.39	0.002	1.46	<0.001	0.98	0.91	1.39	<0.001	1.50	<0.001
	(1.19–1.60)		(1.13–1.71)		(1.26–1.70)		(0.70–1.37)		(1.19–1.62)		(1.27–1.77)
	C-stat 0.69		C-stat 0.70		C-stat 0.70		C-stat 0.71		C-stat 0.70		C-stat 0.71
Multivariable Model 3 (n=1653):
LVET per 10 ms decrease							1.06	0.06	1.03	0.06
							(1.00–1.14)		(1.00–1.06)
							C-stat 0.71		C-stat 0.70
LVET per 10 ms decrease below 350 ms	1.07	0.003	1.07	0.028	1.07	0.009					1.05	0.07
	(1.03–1.11)		(1.02–1.14)		(1.02–1.11)						(1.01–1.10)
	P=0.001		P=0.010		P=0.002						P=0.021
LVET per 10 ms increase above 350 ms	1.00		0.99		1.01						1.02
	(0.96–1.05)		(0.93–1.06)		(0.96–1.05)						(0.97–1.07)
	P=0.93		P=0.82		P=0.75						P=0.52
	C-stat 0.70		C-stat 0.76		C-stat 0.72						C-stat 0.72
LVET/RR per 10% increase	1.13	0.23	0.89	0.43	1.14	0.21	0.91	0.68	1.18	0.14	1.20	0.12
	(0.93–1.38)		(0.67–1.19)		(0.93–1.41)		(0.58–1.43)		(0.95–1.46)		(0.96–1.51)
	C-stat 0.70		C-stat 0.75		C-stat 0.72		C-stat 0.70		C-stat 0.70		C-stat 0.72

^*Using LVET as linear predictor; all other outcomes use LVET as piece-wise linear above/below 350. Model 1 - adjusted for age and gender. Model 2a – Model 1 + hypertension, diabetes, BMI, heart rate, systolic and diastolic BP. Model 2b (Framingham risk score) – Model 1 + total cholesterol, HDL cholesterol, smoking status and systolic blood pressure. Model 3 – Model 2a + FS and LAD. The HR’s are not shown for LVET where the associations have been found to be significantly nonlinear. The Harrel’s c-statistic is displayed for all models below the hazard ratio.

LVET – Left Ventricular Ejection Time; HF – Heart Failure; MI – Myocardial Infarction; BMI – Body Mass Index; FS – Fractional Shortening; LAD – Left Atrium Dimension.

Figure 2: Association between Left Ventricular Ejection Time and risk of incident HF, MI or death, respectively.

Depicting the unadjusted incidence rate of the incident HF (P for overall trend <0.001 and P for non-linearity = 0.002), MI (P for linear and overall trend = 0.024 and P for non-linearity = 0.69) or death (P for overall trend < 0.001 and P for non-linearity < 0.001), respectively (events per 1000 person years) on the y-axis and systolic ejection time (LVET) on the x-axis. The black curve depicts the incidence, with 95% confidence intervals of the estimates.

LVET – Left Ventricular Ejection Time; HF – Heart Failure; MI – Myocardial Infarction.

When assessing LVET as a continuous variable, decreasing LVET was significantly associated with increased risk of all outcomes, particularly for predicting incident HF (Figure 1, 2 and Table 3). After multivariable adjustment for age, gender, hypertension, diabetes, BMI, heart rate, systolic and diastolic BP, LVET remained an independent predictor of all outcomes (Table 3). When adjusting for our conventional measure of systolic and diastolic function, FS, and LAD, LVET remained an independent predictor only of incident HF and the composite outcome (Table 3).

Incremental prognostic information obtained from LVET

The primary risk stratification models used for assessing risk of cardiovascular morbidity and mortality in the general population are currently the Framingham Risk score²⁶ and the SCORE risk chart²⁷. We assessed if adding LVET to the known risk factors obtained from these risk models (SCORE: Age, gender, cholesterol, smoking status and systolic blood pressure and Framingham Risk score: Age, gender, total cholesterol, HDL cholesterol, smoking status and systolic blood pressure) would improve model performance. When adding LVET to the factors from the Framingham Risk score²⁶ or the SCORE risk chart²⁷ the risk models were significantly improved with regard to predicting the composite endpoint (SCORE: c-statistics difference 0.690 (0.671–0.708) vs. 0.678 (0.659–0.696), p<0.001; Framingham Risk score: c-statistics difference 0.695 (0.677–0.713) vs. 0.683 (0.665–0.702), p<0.001), incident HF (SCORE: c-statistics difference 0.713 (0.687–0.738) vs. 0.682 (0.656–0.709), p=0.001; Framingham Risk score: c-statistics difference 0.713 (0.687–0.738) vs. 0.693 (0.667–0.720), p=0.001), and death (SCORE: c-statistics difference 0.706 (0.685–0.727) vs. 0.695 (0.674–0.716), p=0.002; Framingham Risk score: c-statistics difference 0.706 (0.685–0.727) vs. 0.697 (0.676–0.718), p=0.002). Adding LVET to the Framingham Risk score²⁶ or the SCORE risk chart²⁷ did not improve the conventional models for predicting risk of MI (SCORE: c-statistics difference 0.690 (0.645–0.734) vs. 0.682 (0.637–0.727), p=0.11; Framingham Risk score: c-statistics difference 0.712 (0.671–0.753) vs. 0.706 (0.665–0.747), p=0.10).

Similar results were found when adding LVET to the Framingham Risk score as assessed by continuous NRI and IDI. The addition of LVET to models using Framingham risk factors produced significant increases in IDI of approximately +1% and in continuous NRI of approximately +13% with respect to the primary outcome, death, heart failure. Categorical NRI was also significantly improved with respect to the primary outcome and heart failure (+3.0% and +4.8%, respectively; Supplemental Table 1).

Discussion

In the present report we demonstrate the prognostic utility of LVET in a large general population free of cardiovascular disease using long term follow-up. LVET is a significant echocardiographic predictor of cardiovascular morbidity and mortality in a general population with no known cardiovascular disease. Additionally, the LVET is primarily a strong predictor of incident HF, but not MI and to a lesser extent all-cause mortality. LVET provided incremental prognostic information regarding risk of HF and all-cause mortality, but not MI, when added to the Framingham Risk score or the SCORE risk chart.

LVET and cardiovascular outcome

The LVET has previously been demonstrated to be a strong predictor of outcome, especially in patients with ischemic heart disease^8,10,16–19 and primary pulmonary hypertension^12,13. In a previous report, LVET obtained by tissue Doppler imaging (TDI) M-mode was a univariable predictor of a combined outcome of ischemic heart disease, HF, and cardiovascular mortality in a general population²⁹, some of whom had a history of cardiovascular disease. With longer follow-up and therefore more events and higher statistical power, we were able to show that this relationship is mostly due to hospitalization for heart failure.

Weissler and colleagues demonstrated in 1964 that patients with HF had significantly shorter LVET and longer pre-ejection period (PEP = defined as the Q-wave of the ECG – the LVET from the central artery pulse) compared to normal individuals⁷. PEP and LVET are mainly measures of systolic function⁴, and measures of systolic and diastolic function have previously been demonstrated to predict different outcomes³⁰. In the general population, early systolic dysfunction determined by reduced myocardial systolic velocity (TDI s’) or global longitudinal strain have been shown to be a strong predictor of HF^30,31 whereas a reduced diastolic function as determined by a reduced early diastolic relaxation velocity (TDI e’) was a strong predictor of MI³⁰. This is in accordance with our results where our measure of systolic function primarily predicted HF events and not of MI or death (Table 3).

In particular, we found that it was in the low range of LVET that the risk of HF, all-cause mortality, and the combined endpoint increased, whereas a decreasing LVET in the high range wasn’t associated with an increased risk (Table 2 and 3, Figure 1 and 2). This phenomenon has been seen in previous studies^16,29. Furthermore, judging from our results it seems that incident HF primarily drives this range specific pattern of risk obtained from LVET. This is very important because it indicates that there might be a specific cutoff where a decreasing LVET no longer is within the physiological normal range and is a marker of an ailing heart, which if left untreated, might lead to HF. In the present report, the cutoff seems to be somewhere below 350 ms (Figure 1 and 2). However, since there are many methods of obtaining LVET (pulsed wave Doppler of the LVOT³², TDI velocity curves³³, or M-mode TDI^{16,29,34–37}) that lead to different absolute values, further studies are needed to assess the clinically useful cutoffs for each method.

Usefulness of LVET in clinical trials

The LVET is a very easy and fast measure to obtain with high reproducibility^38,39. In contrast to TDI and speckle tracking, LVET is measurable with all conventional echocardiographic machines and software and improves our current risk prediction models. The LVET has therefore been used for several decades to monitor LV contractility and function when testing new medical therapies⁶. In clinical trials, the LVET has previously been demonstrated to improve with medical therapy in patients with ischemic heart disease¹⁰, hypertension¹¹, and HF^14,15. Recently, LVET was used as the primary echocardiographic outcome in a large, multicenter randomized trial⁴⁰ testing a novel drug (omecamtiv mecarbil) that specifically increases the duration of LVET in patients with HF^14,15. Furthermore, the effect of intravenous administration of this drug on LVET was consistent in healthy volunteers⁴¹, chronic HF patients¹⁴, and acute HF patients⁴², as well as the oral study⁴⁰. Even though the LVET has been around for several decades, the usefulness of this simple measure might therefore be revived within the near future.

Limitations

Some limitations have to be noted. First, all participants included were African-Americans, with high BMI and high prevalence of both hypertension and diabetes, which limits the generalizability of our findings to other general populations with another composition. Nevertheless, previous studies comprised primarily of Caucasians have shown that LVET is also predictive of cardiovascular outcomes in general populations with other compositions. However, in these previous studies, the association of LVET with the individual endpoints was not assessed as only the association with composite endpoints was evaluated^29,35. In addition, changes in therapies for hypertension and diabetes mellitus during the last two decades may influence the relevance of the results presented in this report to contemporary populations. Unfortunately, we were not able to assess whether the association between LVET and incident HF was derived from an association with HFpEF or HFrEF, since the ICD-9 codes do not discriminate between the two types of HF admissions. In addition, natriuretic peptides were not measured in the Jackson cohort of the ARIC study. Incident HF occurred more often than MI and less often than death, hence, LVET had greatest power for detecting the association between LVET, HF and death. Furthermore, it should be noted that the Framingham Risk score and the SCORE risk chart both are calibrated to predict risk of ischemic cardiovascular events, which might also explain the finding that LVET did not provide incremental prognostic information regarding risk of MI when added to the known risk factors. Interestingly, despite the fact that these two models were not originally calibrated for incident HF we find that the variables included in the Framingham risk model calibrated well to predict incident HF in the ARIC population as determined by a no-significant difference between predicted and observed number of outcomes (Grønnesby-Borgan X2 statistic of 6.214 and corresponding p= 0.72). We did not adjust for multiple comparisons. Like all other echocardiographic measures of systolic and diastolic function, LVET depends on heart rate^43,44. Hence, physicians should always take heart rate into account when assessing LVET.

Unfortunately, volumetric measurements were not performed as part of the echocardiographic examination of the Jackson cohort of the ARIC study. All measures of cardiac structure and function are therefore based on dimensional measures. Likewise, the pre ejection period (PEP) and the time interval between the late transmitral flow (A wave) and the early transmitral flow (E wave) of the following heart cycle were not measured; hence, we were not able to calculate neither the myocardial performance index nor the PEP/LVET in the present study. The fact that the LVET is a strong predictor of incident HF using echocardiographic machines from 1993 demonstrates the generalizability and how easy it would be to implement in echocardiographic laboratories all around the world despite limited availability of the newest machines and software.

Conclusion

The LVET is a significant predictor of cardiovascular morbidity and mortality in a general population with no known cardiovascular disease. However, the LVET is primarily a strong predictor of incident HF and not MI or all-cause mortality. Furthermore, LVET provides incremental prognostic information regarding risk of HF and all-cause mortality, but not MI, when added to the Framingham Risk score or the SCORE risk chart.