Echocardiographic Predictors of Sudden Cardiac Death: The Atherosclerosis Risk in Communities Study and Cardiovascular Health Study

Abstract

Background

This study assessed the echocardiographic predictors of sudden cardiac death (SCD) within two population-based cohorts.

Methods and Results

Echocardiograms were obtained on 2383 participants (1993–95) from the Atherosclerosis Risk in Communities (ARIC) Study (100% African-American) and 5366 participants (1987–89 and 1994–95) from the Cardiovascular Health Study (CHS). The main outcome was physician-adjudicated SCD. We used Cox proportional hazards models with incident coronary heart disease (CHD) and heart failure as time-dependent covariates to assess the association between echocardiographic variables and SCD, adjusting for Framingham risk score (FRS) variables, CHD, and renal function. Cohort-specific results were meta-analyzed. During a median follow-up of 7.3 years and 13.1 years, 44 ARIC Study and 275 CHS participants had SCD, respectively. In the meta-analyzed results, the adjusted hazard ratios (95% confidence intervals) for predictors of SCD were 3.07 (2.29–4.11) for reduced left ventricular ejection fraction (LVEF); 1.85 (1.36–2.52) for mitral annular calcification; 1.64 (1.07–2.51) for mitral E/A >1.5 and 1.52 (1.14–2.02) for mitral E/A <0.7 (vs mitral E/A 0.7–1.5); 1.30 (1.15–1.48) per one standard deviation (SD) increase in left ventricular mass; and 1.15 (1.02–1.30) per one SD increase in left atrial diameter. A receiver-operating characteristic model for prediction of SCD using FRS variables had a c-statistic of 0.61 for ARIC and 0.67 for CHS; the full multivariable model including all echocardiographic variables had a c-statistic of 0.76 for ARIC and 0.74 for CHS.

Conclusions

In addition to reduced LVEF, we identified other echocardiographic-derived variables predictive for SCD that provided incremental value over clinical risk factors.

Sudden cardiac death (SCD) is a major health concern that accounts for an estimated 300,000 to 375,000 deaths in the United States annually.^1,2 SCD constitutes 50% of deaths due to cardiovascular disease.² Survival from arrest remains poor despite advances in resuscitation. Identification of predisposing factors may be essential in the early intervention and prevention of SCD. A number of methods to predict SCD exist, dating back several decades to include simple, validated tools such as the Framingham risk index.³ While many SCD predictors continue to emerge from the literature, including genetic variants, anatomic assessments using imaging modalities, and various clinical and laboratory data, the cornerstone of SCD risk stratification in patients without a prior cardiac arrest is reduced left ventricular (LV) ejection fraction (LVEF).⁴ Reduced LVEF of any cause is the strongest known predictor of SCD. Despite this, SCD frequently occurs in the absence of reduced LVEF. For example, in the community-based Oregon Sudden Unexpected Death Study (Ore-SUDS), of the 39% with echocardiographic data available, a known reduction in LVEF was present in just 51% of women and 65% of men with SCD.⁵ Information on reduced LVEF alone is therefore not enough to characterize individuals at a higher risk of SCD. The identification of additional SCD risk factors by echocardiogram could help better find those at higher risk of SCD. Given the widespread availability of echocardiography, this could have a substantial impact on the general population.

The objective of this study was to assess the association between echocardiographic variables associated with structural heart changes and SCD among individuals who participated in the ARIC Study and CHS.

Methods

Study population

The ARIC Study has been described elsewhere.⁶ Briefly, the ARIC Study is a prospective population-based investigation of risk factors for atherosclerosis, coronary heart disease (CHD), and stroke. The study includes 15,792 persons aged 45–64 years at baseline (1987–89), randomly chosen from four United States communities: Forsyth County, NC; Jackson, MS; Minneapolis suburbs, MN; and Washington County, MD. In the Jackson, MS cohort, a representative sample was exclusively selected from African-American residents of the city. The ARIC Study cohort members have completed five clinic examinations. Data were first collected in 1987–89, and then again in 1990–92, 1993–95, 1996–98, and 2011–13. Echocardiographic studies for this analysis were performed at the Jackson, MS field site during the third or fourth study visit (1993–96).⁷

The CHS cohort design and recruitment methods have been previously reported.⁸ The CHS consists of a prospectively designed cohort of men and women ≥ 65 years, designed to evaluate cardiovascular risk factors and outcomes in free-living elderly subjects. The original cohort of 5201 patients was selected randomly from Medicare beneficiaries in four US locations: Washington County, MD; Forsyth County, NC; Allegheny County, PA; and Sacramento County, CA. An additional 687 African-Americans were recruited three years later, in 1992 and 1993, to enhance minority representation. A total of 5888 participants (2495 men and 3393 women) were included from 1989–1993. A total of 5683 (5176 from the original cohort and an additional 507 from the African-American cohort) underwent echocardiography. Baseline echocardiographic assessment was performed during the 1989–1990 clinic visit for the initial cohort and during the 1994–1995 clinic visit for the African-American cohort.

A total of 2445 persons in the ARIC Study and the 5683 persons in the CHS underwent echocardiography. We chose echocardiographic predictor variables that are associated with structural changes, were defined similarly in both cohorts, and have known prognostic value. We included participants who had any of the following measurements or assessments available: mitral annular calcification (MAC)(yes/no), aortic sclerosis (yes/no), LVEF (reduced if ejection fraction was <50% in the ARIC Study and <55% in the CHS), left atrium diameter, aortic root diameter, LV mass, LV mass index, interventricular septal thickness (IVST), posterior wall thickness (PWT), relative wall thickness (RWT), LV internal dimensions in end diastole (LVIDd) and systole (LVIDs), and transmitral peak E-wave and A-wave velocities. In the CHS, roughly a third of patients were missing M-mode measurements, and therefore we could only calculate LV mass in 63.5% of the CHS cohort. Older age, chronic obstructive pulmonary disease, and obesity were associated with missing M-mode data. The sample size for our primary analysis was 2383 ARIC Study and 5366 CHS participants, which included 95.3% of those with echocardiography.

All study participants provided informed consent at baseline and follow-up exams. The ARIC Study and CHS were approved by Institutional Review Boards at participating institutions.

Echocardiographic measurements

The designs of the echocardiographic protocols for the ARIC Study⁷ and CHS^9,10 were described previously. The IVST, PWT, LVIDd, LVIDs, left atrium diameter, and aortic root diameter were obtained using M mode according to the conventions established by the American Society of Echocardiography at the time of the index examination. LV mass was derived using the formula described by Devereaux and associates using the M mode data.¹¹ The ratio of PWT and IVST versus LVIDd determined relative wall thickness. The criterion for MAC was similar to previous studies.^12–14 MAC was defined as an echo-dense structure located at the atrioventricular groove and posterior mitral leaflet, visualized in multiple views in both cohorts. Aortic valve sclerosis was similarly identified as focal or diffuse aortic cusp thickening, stiffness, and/or increased echogenicity (calcification) with normal aortic cusp excursion and a peak trans-aortic valve flow velocity <2.0 m/s in both cohorts.

Mitral inflow velocity was measured by pulsed-wave Doppler from the apical 4-chamber view with the sample volume positioned at the tips of the mitral leaflets. The inflow variables analyzed were 1) peak Doppler E wave velocity and 2) early (E wave) to late (A wave) ratio (E/A). Additional specific information about the echocardiography studies performed in the 2 cohorts is outlined below.

Echocardiogram Performance and Reading Protocol

Briefly, for the ARIC Study participants, images were acquired using the Acuson XP128/10c echocardiography machine (GPS medical). A single cardiologist performed all echocardiographic readings. LV systolic function was considered reduced if LVEF was <50%. LVEF was derived semi-quantitatively using a modified Quinones technique and visual assessment of the LV apex.

In the CHS cohort, 2-dimensional images were recorded on super VHS tape following a standard protocol with the use of a cardiac ultrasound machine (model SSH-160A, Toshiba, Tustin, California). Two trained independent readers, both unaware of the participants’ clinical information, read images at a core laboratory. Global LV systolic function was qualitatively assessed from the 2 dimensional echocardiogram images. Left ventricular systolic function was subjectively categorized as normal (EF≥55%), borderline (EF<55 and ≥45), or abnormal (EF<45%), with 94% inter-reader agreement and 98% intrareader agreement of paired studies (eMethods). In this study, left ventricular systolic function was considered reduced if LVEF was <55%. Representative images from the CHS are shown in Supplemental Figures 1–4.

Outcomes ascertainment

In the ARIC Study and CHS cohorts, respective ARIC Study and CHS events committees classified cardiovascular deaths. A separate review of the CHD deaths was conducted to identify SCD events, and included review of death certificates, and when available, autopsy reports, next-of-kin interviews, and questionnaires to decedents’ physicians. The reviewers were blinded to the echocardiographic data. The primary outcome, SCD, was similarly defined in both the ARIC Study and CHS: a sudden pulseless condition presumed to be due to a ventricular tachyarrhythmia in a previously stable individual without evidence of a noncardiac cause of cardiac arrest. All SCD cases occurred outside the hospital or in the emergency department, and the individuals could not have a life-threatening noncardiac comorbidity or be under hospice or nursing home care. For unwitnessed SCDs, the participant must have been seen within 24 hours of the arrest in a stable condition and without evidence of a noncardiac cause of cardiac arrest.

In the ARIC Study, all CHD deaths that occurred by December 31, 2001, were reviewed by a panel of 5 physicians to identify SCD events. Each event was independently classified by 2 physicians. If there was disagreement, a third investigator reviewed the event to provide a final classification. After review of available data, CHD deaths were classified as definite sudden arrhythmic death, possible sudden arrhythmic death, not sudden arrhythmic death, or unclassifiable.¹⁵ For the present analysis, SCD was defined as definite or possible sudden arrhythmic death in the ARIC Study.

In the CHS, all CHD deaths through December 31, 2006, were reviewed by a cardiologist to classify SCD cases. A blinded second physician review of a random sample of 70 of these death records showed an 88% inter-reviewer agreement and κ = 0.74 for SCD. Both of these physicians also participated on the ARIC Study SCD review panel. After review of available data in the CHS, CHD deaths were classified as definite, possible, or not SCD.¹⁶ For the present analysis, the CHS definition of SCD included definite and possible SCD.

In both cohorts, the secondary outcome, non-sudden cardiovascular (CVD) deaths (NSCD), was defined as CVD death not meeting SCD criteria. For the ARIC Study cohort, International Classification of Diseases codes were used from patient death certificates to identify CVD deaths.

Assessment of other covariates

For this analysis we used covariates measured at the time of echocardiogram. Definitions of the covariates are detailed in the Supplemental Methods.

Statistical analysis

Means (standard deviation [SD]) for continuous variables and counts with percentages for categorical variables were calculated. Person-years at risk were calculated from the date of the echocardiographic exam until the date of SCD or NSCD, other death, loss to follow-up, or end of follow-up, whichever occurred first.

Cohort-specific analyses were conducted. Cox proportional-hazards models were used to calculate hazard ratios (HRs) and 95% confidence intervals (CIs) of SCD by echocardiographic variables. An initial model adjusted for age, sex, and in the CHS, race. A second model additionally adjusted for Framingham risk score and prevalent CHD. A third model also adjusted for renal function (estimated glomerular filtration rate [eGFR]). Finally, a fourth model also adjusted for incident CHD and heart failure as time-dependent covariates and potential mediators of the association of echocardiographic variables with SCD. Cohort-specific analyses were combined using fixed-effects meta-analysis. Between-cohort heterogeneity was determined calculating the I-squared statistic and the Cochrane’s Q statistic for heterogeneity. Low levels of the I-squared statistic and non-significant Q statistic indicate lack of evidence of heterogeneity between both cohorts. The combined associations were considered the primary results. A similar set of analyses was conducted for NSCD.

Receiver-operating characteristic curve analysis was constructed to determine the incremental value of incorporating echocardiographic variables to Framingham risk score variables for the prediction of SCD. A new value was built for each individual echocardiographic variable added to the Framingham risk score. Also, a new value was built for all echocardiographic variables added to the Framingham risk score.

Commercial software was used for statistical analysis of ARIC Study data (SAS, version 9.2; SAS Institute Inc.) and CHS data (SPSS, version 16; SPSS, Inc; and Stata, version 11.2; StataCorp). All P values reported were 2-sided, and were considered statistically significant if <0.05. However, the focus of our primary analysis was in estimation of HRs and their 95%CIs, rather than statistical significance.

Results

The incidence of SCD in the ARIC Study was 2.59 (1.91–3.44) per 1,000 person-years and in the CHS was 4.38 (3.89–4.93) per 1,000 person-years over a median (interquartile range) follow-up of 7.3 (1.4) and 13.1 (8.3) years, respectively. Table 1 lists the baseline characteristics of both cohorts. The CHS cohort was older with a mean baseline age of 72.9 years versus 58.8 years in the ARIC Study. All participants in the ARIC Study were African-American, compared to 12.3% in the CHS. The prevalence of anti-arrhythmic drug therapy was low in either cohort.

Table 1.

Baseline characteristics in the ARIC Jackson field center, 1993–1995, and CHS, 1989–1995, stratified by later SCD occurrence

Variable	ARIC			CHS
	Total sample (n=2383)	No SCD (n=2339)	SCD (n=44)	Total sample (n=5366)	No SCD (n=5091)	SCD (n=275)
Age, years	58.8 (5.7)	58.8 (5.7)	61.6 (5.6)	72.9 (5.6)	72.9 (5.5)	73.3 (5.4)
Female, %	64	65	55	58	59	40
African-American, %	100	100	100	12	12	15
Current smoker, %	20	20	30	12	12	15
Hypertension, %	60	60	77	58	57	67
Diabetes, %	24	24	30	15	15	25
Total cholesterol ≥ 240 mg/dL, %	18	19	14	21	22	20
HDL cholesterol < 60 mg/dL, %	65	65	80	69	68	81
Prevalent CHD, %	5	4	23	19	18	35
eGFR mL/min/1.73 m2	101 (19)	102 (19)	91 (18)	79 (23)	79 (23)	77 (23)
Beta-blockers, %	9	9	11	13	13	16
Anti-arrhythmics, %	0.6	0.5	2	3	3	6
Calcium channel blockers, %	16	16	27	13	13	16

ARIC=Atherosclerosis Risk in Communities Study; CHD=coronary heart disease; CHS=Cardiovascular Health Study; eGFR=estimated glomerular filtration rate; HDL=high density lipoprotein; SCD=sudden cardiac death

Values correspond to either means (standard deviation) or percentages

Table 2 shows the baseline echocardiographic variables in both cohorts. Participants in the ARIC Study had a greater mean LV mass and a slightly greater mitral peak E velocity; the CHS had more with reduced LVEF, presence MAC, aortic sclerosis, and a mitral E/A less than 0.7 m/s.

Table 2.

Echocardiographic variables at baseline, ARIC, 1993–1998, and CHS 1989–1995, stratified by later SCD occurrence

Variable	ARIC				CHS
	N	Total Sample	No SCD (n=2339)	SCD (n=44)	N	Total Sample	No SCD (n=5091)	SCD (n=275)
Mitral annular calcification (%)	2368	5	5	14	3757†	42	41	58
Reduced LV ejection fraction (%)	2373	3	2	16	5091†	9	8	24
Aortic sclerosis (%)	2368	5	5	13	3657†,*	45	45	49
Left atrium diameter, cm	1931	4 (0.6)	4 (0.5)	4 (0.8)	5192	4 (0.7)	4 (0.7)	4 (0.7)
Aortic root diameter, cm	2149	3 (0.4)	3 (0.4)	3 (0.5)	5203	3 (0.5)	3 (0.5)	3 (0.5)
LV mass index, g/m^2.7	1815	50 (16)	50 (16)	60 (26)	3611†	39 (13)	39 (12)	44 (14)
Mitral E/A
< 0.7	184	8%	8%	13%	1033	20%	20%	27%
0.7 to 1.5	1847	85%	85%	79%	3854	75%	75%	64%
> 1.5	150	7%	7%	8%	303	6%	6%	9%
Mitral Peak E, m/s	2187	0.77 (0.2)	0.77 (0.2)	0.84 (0.3)	5201	0.72 (0.2)	0.72 (0.2)	0.72 (0.2)

ARIC=Atherosclerosis Risk in Communities Study; CHD=coronary heart disease; CHS=Cardiovascular Health Study; SCD=sudden cardiac death

Values correspond to mean (standard deviation) or percentage

^†Reduced ‘n’ to between 3550 to 3612

^*No data available for second African American cohort

Table 3 shows the cohort specific data. In a multivariable model adjusted for potential confounders (Table 3, Model 3), reduced LVEF, presence of MAC, and increased LV mass were associated with SCD in both cohorts. In the ARIC Study, increased aortic root diameter and mitral peak E velocity were also associated with SCD. In the CHS, increased left atrial diameter, and mitral E/A <0.7 or >1.5 were also associated with SCD. With the addition of CHD and heart failure as time-dependent covariates (Table 3, Model 4), only increased LV mass and mitral peak E velocity remained associated with SCD in the ARIC Study, but there was little effect on the associations in the CHS.

Table 3.

SCD incidence rate and hazard ratio (95% confidence interval) by echocardiographic variables, ARIC, 1993–2002, and CHS 1989–2002

	Mitral annular calcification
	ARIC			CHS
	No	Yes	p-value	No	Yes	p-value
SCD cases	37	6		67	91
Person-years	16118	796		29696	20088
SCD Incidence rate (95% CI)*	2.30 (1.64 – 3.13)	7.54 (3.13 – 15.5)		2.26 (1.78 – 2.87)	4.53 (3.69 – 5.56)
Model 1: HR (95% CI)	1(REF)	2.78 (1.16 – 6.63)	0.02	1(REF)	2.07 (1.50 – 2.85)	<0.001
Model 2: HR (95% CI)	1(REF)	2.71 (1.14 – 6.45)	0.02	1(REF)	2.07 (1.51 – 2.84)	<0.001
Model 3: HR (95% CI)	1(REF)	2.56 (1.08 – 6.09)	0.03	1(REF)	2.07 (1.51 – 2.85)	<0.001
Model 4: HR (95% CI)	1(REF)	1.78 (0.73 – 4.35)	0.20	1(REF)	1.86 (1.34 – 2.58)	<0.001
	Aortic Sclerosis
	ARIC			CHS
	No	Yes	p-value	No	Yes	p-value
SCD cases	37	6		78	74
Person-years	16041	879		27318	21277
SCD Incidence rate (95% CI)*	2.31 (1.65 – 3.14)	6.83 (2.84 – 14.07)		2.86 (2.29 – 3.57)	3.48 (2.77 – 4.37)
Model 1: HR (95% CI)	1(REF)	2.29 (0.95 – 5.51)	0.06	1(REF)	1.21 (0.88 – 1.67)	0.25
Model 2: HR (95% CI)	1(REF)	2.40 (1.01 – 5.72)	0.05	1(REF)	1.27 (0.93 – 1.75)	0.14
Model 3: HR (95% CI)	1(REF)	2.16 (0.90 – 5.16)	0.08	1(REF)	1.28 (0.93 – 1.76)	0.13
Model 4: HR (95% CI)	1(REF)	1.88 (0.78 – 4.53)	0.16	1(REF)	1.09 (0.79 – 1.53)	0.59
	Reduced LV ejection fraction
	ARIC			CHS
	No	Yes	p-value	No	Yes	p-value
SCD cases	36	7		208	67
Person-years	16629	341		58667	4143
SCD Incidence rate (95% CI)*	2.16 (1.54 – 2.96)	20.53 (9.15 – 40.31)		3.55 (3.10 – 4.06)	16.17 (12.7 – 20.6)
Model 1: HR (95% CI)	1(REF)	8.13 (3.56 – 18.56)	<0.0001	1(REF)	3.89 (2.93 – 5.16)	<0.001
Model 2: HR (95% CI)	1(REF)	5.06 (2.06 – 12.45)	0.0004	1(REF)	3.26 (2.44 – 4.37)	<0.001
Model 3: HR (95% CI)	1(REF)	4.25 (1.70 – 10.63)	0.002	1(REF)	3.24 (2.42 – 4.33)	<0.001
Model 4: HR (95% CI)	1(REF)	1.95 (0.76 – 4.98)	0.16	1(REF)	3.22 (2.37 – 4.37)	<0.001

	Left atrium diameter
	ARIC tertiles (cm)			CHS tertiles (cm)
	<3.62	3.62 – 4.05	>4.05	<3.60	3.60 – 4.13	>4.13
SCD cases	9	9	17	71	80	112
Person-years	4580	4572	4634	20881	20997	19032
SCD Incidence rate (95% CI)*	1.97 (0.97 – 3.59)	1.97 (0.97 – 3.59)	3.67 (2.22 – 5.74)	3.40 (2.70 – 4.29)	3.81 (3.06 – 4.74)	5.89 (4.89 – 7.08)
Model 1: HR (95% CI)	1 (ref.)	1.00 (0.40 – 2.52)	1.81 (0.81 – 4.07)	1 (ref.)	1.05 (0.77 – 1.45)	1.46 (1.08 – 1.98)
Model 2: HR (95% CI)	1 (ref.)	0.97 (0.38 – 2.44)	1.50 (0.69 – 3.54)	1 (ref.)	1.03 (0.75 – 1.42)	1.34 (0.99 – 1.81)
Model 3: HR (95% CI)	1 (ref.)	0.93 (0.65 – 3.37)	1.48 (0.65 – 3.37)	1 (ref.)	1.03 (0.75 – 1.41)	1.33 (0.98 – 1.80)
Model 4: HR (95% CI)	1 (ref.)	0.91 (0.36 – 2.41)	1.40 (0.61 – 3.21)	1 (ref.)	1.06 (0.76 – 1.48)	1.27 (0.92 – 1.74)

	Aortic root diameter
	Aric tertiles (cm)			CHS tertiles (cm)
	<2.92	2.92 – 3.29	>3.29	<2.96	2.96 – 3.35	>3.35
SCD cases	8	12	22	73	81	110
Person-years	5195	4984	5146	21487	20747	18791
SCD Incidence rate (95% CI)*	1.54 (0.73 – 2.91)	2.41 (1.31 – 4.08)	4.28 (2.76 – 6.36)	3.40 (2.70 – 4.27)	3.90 (3.14 – 4.85)	5.85 (4.86 – 7.06)
Model 1: HR (95% CI)	1 (ref.)	1.43 (0.58 – 3.53)	2.34 (0.97 – 5.62)	1 (ref.)	0.94 (0.68 – 1.30)	1.01 (0.72 – 1.42)
Model 2: HR (95% CI)	1 (ref.)	1.42 (0.58 – 3.49)	2.59 (1.15 – 5.84)	1 (ref.)	0.98 (0.72 – 1.35)	1.23 (0.90 – 1.68)
Model 3: HR (95% CI)	1 (ref.)	1.38 (0.56 – 3.38)	2.41 (1.06 – 5.45)	1 (ref.)	0.98 (0.71 – 1.35)	1.21 (0.89 – 1.66)
Model 4: HR (95% CI)	1 (ref.)	1.34 (0.55 – 3.29)	2.12 (0.92 – 4.85)	1 (ref.)	0.99 (0.71 – 1.37)	1.13 (0.82 – 1.57)

	LV mass index
	ARIC tertiles (g/m^2.7)			CHS tertiles (g/m^2.7)
	<41.4	41.4 – 53.1	>53.1	<33.07	33.07 – 42.20	>42.20
SCD cases	7	9	16	36	51	79
Person-years	4278	4318	4327	15391	15002	13197
SCD Incidence rate (95% CI)*	1.64 (0.73 – 3.21)	2.08 (1.03 – 3.80)	3.70 (2.20 – 5.86)	2.34 (1.69 – 3.24)	3.40 (2.58 – 4.47)	5.99 (4.80 – 7.46)
Model 1: HR (95% CI)	1 (ref.)	1.24 (0.46 – 3.34)	2.09 (0.86 – 5.09)	1 (ref.)	1.48 (0.96 – 2.26)	2.41 (1.62 – 3.58)
Model 2: HR (95% CI)	1 (ref.)	1.12 (0.41 – 3.01)	1.58 (0.62 – 4.01)	1 (ref.)	1.38 (0.90 – 2.12)	2.02 (1.35 – 3.01)
Model 3: HR (95% CI)	1 (ref.)	1.11 (0.41 – 2.99)	1.49 (0.58 – 3.79)	1 (ref.)	1.39 (0.91 – 2.14)	2.03 (1.36 – 3.02)
Model 4: HR (95% CI)	1 (ref.)	1.21 (0.45 – 3.26)	1.40 (0.56 – 3.54)	1 (ref.)	1.26 (0.81 – 1.96)	1.69 (1.12 – 2.58)

	Mitral E to A
	ARIC (m/s)			CHS (m/s)
	<0.70	0.70 – 1.5	>1.5	<0.70	0.70 – 1.5	>1.5
SCD cases	5	30	3	71	170	23
Person-years	1270	13294	1030	10106	47698	3170
SCD Incidence rate (95% CI)*	3.94 (1.49 – 8.63)	2.26 (1.55 – 3.18)	2.91 (0.81 – 7.77)	7.03 (5.57 – 8.87)	3.56 (3.07 – 4.14)	7.26 (4.82 – 10.92)
Model 1: HR (95% CI)	1.40 (0.43 – 4.71)	1 (ref.)	1.43 (0.43 – 4.71)	1.84 (1.39 – 2.46)	1 (ref.)	1.94 (1.25 – 3.00)
Model 2: HR (95% CI)	1.15 (0.42 – 3.11)	1 (ref.)	1.35 (0.41 – 4.49)	1.84 (1.39 – 2.44)	1 (ref.)	1.82 (1.18 – 2.83)
Model 3: HR (95% CI)	0.97 (0.36 – 2.66)	1 (ref.)	1.42 (0.43 – 4.69)	1.81 (1.37 – 2.40)	1 (ref.)	1.78 (1.15 – 2.77)
Model 4: HR (95% CI)	0.24 (0.08 – 0.73)	1 (ref.)	0.91 (0.33 – 2.56)	1.73 (1.29 – 2.32)	1 (ref.)	1.85 (1.16 – 2.96)

	Mitral Peak E
	ARIC (m/s)		CHS (m/s)
	Continuous, per 1 SD	p-value+	Continuous, per 1 SD	p-value+
Model 1: HR (95% CI)	1.49 (1.14 – 1.96)	0.004	1.11 (0.98 – 1.26)	0.09
Model 2: HR (95% CI)	1.37 (1.06 – 1.77)	0.02	1.04 (0.92 – 1.17)	0.55
Model 3: HR (95% CI)	1.39 (1.08 – 1.79)	0.01	1.04 (0.92 – 1.18)	0.52
Model 4: HR (95% CI)	1.29 (1.01 – 1.66)	0.04	1.04 (0.92 – 1.18)	0.51

ARIC=Atherosclerosis Risk in Communities Study; CHD=coronary heart disease; CHS=Cardiovascular Health Study; CVD=cardiovascular disease; eGFR=estimated glomerular filtration rate; HDLc=high density lipoprotein cholesterol; LV=left ventricular; SCD=sudden cardiac death

^*Crude incidence rates per 1000 person-years

Refer to Supplemental Table 2 for continuous variable data.

Model 1: Cox proportional hazards model adjusted for age, sex, **race (in CHS only)

Model 2: Cox proportional hazards model adjusting for Framingham risk score as one variable: (sex, age, total cholesterol, HDLc, blood pressure, diabetes and smoking), and also prevalent CHD

Model 3: Model 2 + eGFR

Model 4: Model 3 + time-dependent heart failure, time-dependent CHD

The association of echocardiographic variables with NSCD was also assessed (Supplemental Table 1).

A receiver-operating characteristic model for prediction of SCD using FRS variables had a c-statistic of 0.61 (95%CI 0.53–0.69) for ARIC and 0.67 (0.64–0.70) for CHS; the full multivariable model including all echocardiographic variables had a c-statistic of 0.76 for ARIC (0.66–0.87) and 0.74 for CHS (0.69–0.75) (Table 4).

Table 4.

C-statistic of sudden cardiac death with echocardiographic variables added to the Framingham risk score

	ARIC C-statistic (95% CI)	CHS C-statistic (95% CI)
Framingham risk score	0.608 (0.526–0.690)	0.669 (0.638–0.700)
FRS + Univariate: (each variable added to the score separately)
Mitral annular calcification	0.648 (0.567–0.729)	0.677 (0.633–0.721)
Reduced LV ejection fraction	0.659 (0.573–0.744)	0.719 (0.688–0.750)
Aortic sclerosis	0.620 (0.535–0.705)	0.671 (0.628–0.714)
Left atrium diameter, continuous	0.608 (0.522–0.694)	0.669 (0.637–0.702)
Left atrium diameter, tertiles	0.619 (0.537–0.702)	0.669 (0.637–0.701)
Aortic root diameter, continuous	0.682 (0.609–0.756)	0.666 (0.635–0.697)
Aortic root diameter, tertiles	0.637 (0.564–0.710)	0.666 (0.635–0.697)
LV mass index, continuous	0.648 (0.542–0.755)	0.687 (0.645–0.728)
LV mass index, tertiles	0.623 (0.538–0.707)	0.670 (0.628–0.712)
Mitral E/A, continuous	0.629 (0.542–0.715)	0.666 (0.634–0.698)
Mitral E/A, categories	0.633 (0.553–0.713)	0.671 (0.639–0.703)
Mitral Peak E, continuous	0.644 (0.552–0.735)	0.668 (0.635–0.700)
FRS + Multivariate: (all variables added to the score)	0.764 (0.657–0.870)	0.741 (0.687–0.749)

The meta-analyzed data for the ARIC Study and CHS, using Model 4 adjustments, found several echocardiographic variables to have significant associations with SCD. Reduced LVEF showed a threefold risk of SCD, presence of MAC a twofold risk of SCD, and mitral E/A >1.5 or <0.7 a one-and-a-half fold risk of SCD. Both LV mass and increased left atrial diameter were also statistically significant (Figure).

Figure.

Forest plot of the meta-analyzed and risk-adjusted hazard ratios for SCD in the ARIC Study and CHS, using CHD and heart failure as time-dependent variables (Model 4)

For NSCD, the meta-analyzed results using Model 4 adjustments had HRs (95%CI) of 1.77 (1.46–2.13) for reduced LVEF; 1.70 (1.47–1.97) for mitral E/A <0.7; 1.53 (1.20–1.96) for mitral E/A>1.5; 1.45 (1.23–1.71) for MAC; and 1.44 (1.22–1.71) for aortic sclerosis. LV mass and LAD had smaller, albeit significant, associations with NSCD (Supplemental Figure 5).

In addition, a receiver-operating characteristic model for prediction of SCD using FRS variables had a c-statistic of 0.61 for ARIC and 0.67 for CHS; the full multivariable model including all echocardiographic variables had a c-statistic of 0.76 for ARIC and 0.74 for CHS.

To assure the validity of the qualitative EF assessment in the CHS, we performed a sensitivity analysis in which we substituted EF with fractional shortening, a quantitative assessment of left ventricular performance. This resulted in only a minimal change in the cumulative c-statistic to 0.73 (0.68–0.79) (Supplemental Table 3).

Discussion

This large population-based study evaluated the association of multiple echocardiographic variables with SCD, with findings having several important clinical implications. Based on the meta-analyzed results from the ARIC Study and CHS cohorts, recognition should be given to MAC, mitral E/A >1.5 or <0.7, increased LV mass, and increased left atrial diameter, in addition to reduced LVEF, as potential markers of SCD in the general population. Of the previously accepted echocardiographic predictors of SCD (reduced LVEF and LV hypertrophy),^4,17 only reduced LVEF has been studied in randomized controlled trials investigating the use of implantable cardioverter defibrillators (ICDs), which reduced all-cause mortality in persons having ischemic^18–20 or non-ischemic²⁰ LV dysfunction. Yet many SCDs occur in patients not having recognized LV dysfunction. In a population-based study in the Netherlands, among the 41% with prior echocardiographic data available, just 50% had an LVEF ≤50%.²¹ In the community-based Ore-SUDS study, of the 39% with previous echocardiographic data, a reduction in LVEF was present in 51% of women and 65% of men. In these studies, it is possible the overall prevalence of LV dysfunction would have been different had the entire population been screened with echocardiograms. In our study, just 16.3% of ARIC Study and 24.4% of CHS participants with SCD had reduced LVEF at the time of their baseline echocardiogram. Based on current guidelines for ICD placement, which is largely based on reduced LVEF, many of the subjects in these studies would not have had an ICD indication. These findings suggest a need to identify additional predictors of SCD besides reduced LVEF.

In this analysis, MAC was associated with increased SCD incidence after adjusting for risk factors. MAC has been associated with increased mortality, CHD, congestive heart failure, and ischemic stroke.^{12–14,22,23} Although the mechanism is not well understood, MAC has been considered to be a marker of atherosclerosis¹³ and perhaps even a form of atherosclerosis.^24,25 It is believed that MAC and atherosclerosis are related through a number of coronary risk factors.^13,14 Certain conduction abnormalities are higher in patients with MAC and conduction delay is a risk factor for SCD.² Previous reports have implicated MAC as a cause of SCD occurring in the absence of acute ischemia.²⁶ This is the first population based cohort study, to our knowledge, showing a positive association between MAC and SCD.

Several studies support an association between LVH and CVD,^{2,17,23,27–29} and in this study we found that LVH, as measured by increased LV mass, predicted greater incidence of SCD after adjustments for risk factors and CHD. While electrocardiographic (ECG) detection of LVH is also associated with increased SCD risk,³⁰ echocardiographic measurement of LV mass is a more accurate and reliable tool than ECG for measuring LVH.^31–33 A plausible explanation for SCD in those with increased LV mass is that it can precede LV dysfunction.³⁴ Other postulated mechanisms include myocardial fibrosis, a component in the development of LVH, known to facilitate reentrant ventricular arrhythmias.^35,36 Fibrosis and LVH may lead to neurohormonal activation and further fibrosis due to angiotensin II.³⁷ Neurohormonal activation and autonomic dysregulation may also directly predispose to arrhythmias. Finally, prolonged ventricular repolarization may occur in LVH, increasing the susceptibility to ventricular fibrillation.³⁸

A prior study from the CHS (age ≥65 years) showed that increasing aortic root diameter was a modest predictor of increased CVD mortality,³⁹ but a Chinese cohort did not show a significant association between greater aortic root diameter and CVD mortality after age adjustment.⁴⁰ Investigations on aortic root diameter and CVD mortality are otherwise sparse in the literature, and there are no studies examining SCD. A potential explanation for SCD in those with aortic root diameter is its association with both increased LV mass and LV dysfunction.^41,42 While a previous CHS analysis showed aortic valve sclerosis to confer a 50% increased risk of cardiovascular morbidity and mortality, it was not associated with increased SCD in either the ARIC Study or CHS cohorts.

This is the first study to find an association between left atrium diameter and SCD, while a prior study found it to predict first clinical cardiovascular events.^23,43,44 Given the well-known association between left atrium diameter and atrial fibrillation, and the latter shown to be associated with SCD,⁴⁵ it is possible that atrial fibrillation could explain this association between SCD and left atrium diameter. However, left atrial diameter increases with increasing diastolic load, and is therefore also a marker of diastolic heart failure.

SCD is common in diastolic heart failure (i.e. heart failure with preserved ejection fraction). SCD accounted for 28% of all deaths in the CHARM-Preserved study⁴⁶ (candesartan versus placebo, LVEF>40%) and 26% of all deaths in the I-PRESERVE study⁴⁷ (irbesartan versus placebo, LVEF ≥45%). Investigators have therefore sought ways of predicting SCD in this group.⁴⁸ Previous CHS data showed peak E velocity to predict incident CHF and the mitral E/A ratio to predict incident CHF in a U-shaped manner.⁴⁹ The meta-analyzed ARIC Study and CHS data found mitral E/A <0.7 or >1.5 to be associated with increased SCD incidence. However, the individual ARIC Study and CHS results were discrepant, with the ARIC Study cohort demonstrating a lower SCD risk with a mitral E/A <0.7. Without other important diastolic parameters such as left atrial volume and mitral annular tissue velocities, along with the low number of SCD in the ARIC Study, it may be difficult to interpret the significance of this finding.

Finally, we should also acknowledge that traditional risk factors help identify SCD, and that clinical trials have shown that the treatment of risk factors will reduce CHD mortality, much of which is SCD.⁵⁰ A population-based, preventive strategy aimed at avoiding or reducing risk factors would thus reduce the public health burden of SCD.

Strengths

The ARIC Study and CHS are two large population-based cohorts addressing CVD in the United States with prospective information on echocardiographic risk markers and SCD. In addition, there was a long follow-up, inclusion of non-white participants, extensive measurement of covariates, and physician adjudication of all SCD cases. Finally, the identified echocardiographic risk markers are easily obtainable, and our results may be generalizable to daily practice.

Limitations

Our study has several limitations. The meta-analysis weighs more heavily towards CHS given the ratio of SCD of about 6:1 between CHS and ARIC. In addition, the power to detect heterogeneity between only two studies is low. The power to discriminate between different echocardiographic covariates was limited by relatively low numbers of SCD. We did not adjust for possible intercorrelated echocardiographic covariates to determine if some were superfluous. For example, when calculating the risk of SCD for reduced LVEF, we did not adjust for LV mass, which is a predictor of reduced LVEF.³⁴ Changes in echocardiographic variables might have occurred during follow-up, which would have led to misclassification and potential bias in the hazard ratios. The same might be true for CHD and HF status, though we adjusted for these as time-dependent covariates. Variables such as dyssynchrony, wall motion abnormalities, left atrial volume, and other diastolic parameters were not available. Importantly, the SCD variable was an epidemiological classification and, in the event of unwitnessed SCD, it was impossible to determine whether there was underlying ischemia. In most cases, we did not know the degree of underlying coronary artery disease at the time of death or whether the SCD was solely arrhythmic.

Conclusions

This study suggests that in a middle-aged to elderly population, reduced LVEF, MAC, mitral E/A >1.5 or <0.7, increased LV mass, and left atrial diameter are associated with an increased risk of SCD. In addition, these echocardiographic variables provided incremental prognostic value over clinical risk factors. These readily measurable variables could help identify patients who would benefit from preventive or therapeutic interventions and, in turn, reduce death rates. However, the clinical utility of their measurement and any interventions would need to be assessed in randomized clinical trials.

Clinical Perspective.

Sudden cardiac death (SCD) in the community continues to be a major public health problem. In this study we assessed baseline echocardiographic variables from 2383 participants from the Atherosclerosis Risk in Communities Study and 5366 participants from the Cardiovascular Health Study. We found several echocardiographic predictors of SCD. These included reduced left ventricular ejection fraction, mitral annular calcification, mitral E/A >1.5 or <0.7, increased left ventricular mass, and increased left atrial diameter. Furthermore, these echocardiographic predictors provided incremental value over clinical risk factors in predicting SCD. We believe clinicians can use this information to better understand who may be at risk of SCD.