- Email: [email protected]
M-Mode echocardiographic predictors of six- to seven-year incidence of coronary heart disease, stroke, congestive heart failure, and mortality in an elderly cohort (the cardiovascular health study)
M-Mode Echocardiographic Predictors of Six- to Seven-Year Incidence of Coronary Heart Disease, Stroke, Congestive Heart Failure, and Mortality in an Elderly Cohort (The Cardiovascular Health Study) Julius M. Gardin, MD, Robyn McClelland, PhD, Dalane Kitzman, MD, Joao A.C. Lima, MD, William Bommer, MD, H. Sidney Klopfenstein, MD, PhD, Nathan D. Wong, PhD, Vivienne-Elizabeth Smith, MD, and John Gottdiener, MD Previous studies have identified a number of echocardiographic variables that predict cardiovascular disease (CVD) events and mortality, but have not focused on a large elderly cohort. The purpose of this study was to determine whether M-mode echocardiographic variables predicted all-cause mortality, incident coronary heart disease (CHD), congestive heart failure (CHF), and stroke in a large prospective, multicenter, populationbased study. In the Cardiovascular Health Study, a biracial cohort of 5,888 men and women (mean age 73 years) underwent 2-dimensional M-mode echocardiographic measurements of left ventricular (LV) internal dimensions, wall thickness, mass and geometry, as well as measurement of left atrial dimension and assessment for mitral annular calcium. Participants were followed for 6 to 7 years for incident events; analyses excluded subjects with prevalent disease. One or more echocardiographic measurements were independent predictors of all-cause mortality and incident CHD, CHF, and
stroke. After adjustment for anthropometric and traditional CVD risk factors, LV mass was significantly related to incident CHD, CHF, and stroke. The highest quartile of LV mass conferred a hazards ratio of 3.36, compared with the lowest quartile, for incident CHF. Furthermore, incident CHF-free survival was significantly lower for participants with LV mass in the highest versus the 2 lowest quartiles (86% vs 97%, respectively, at 2,500 days). Eccentric and concentric LV hypertrophy, respectively, conferred adjusted hazards ratios, compared with normal LV geometry, of 2.05 and 1.61 for incident CHD, and 2.95 and 3.32 for incident CHF. Thus, in an elderly biracial population, selected 2-dimensional Mmode echocardiographic measurements were important markers of subclinical disease and conferred independent prognostic information for incident CVD events, especially CHF and CHD. 䊚2001 by Excerpta Medica, Inc. (Am J Cardiol 2001;87:1051–1057)
revious studies have identified a number of transthoracic M-mode echocardiographic variables P with independent prognostic value for cardiovascular
cohort study of elderly men and women sponsored by the National Heart, Lung, and Blood Institute. This study focuses on a 6- to 7-year follow-up period, examining the relation of M-mode echocardiographic variables, including left ventricular (LV) mass and its component variables, LV geometric subtypes (including concentric and eccentric hypertrophy and concentric remodeling), left atrial dimension, and mitral annular calcium, to incident coronary heart disease (CHD), congestive heart failure (CHF), stroke, and (cardiac and overall) mortality.13,14
disease (CVD) events and overall mortality. However, these studies have generally not focused on an elderly population-based cohort.1–12 The Cardiovascular Health Study (CHS) is a longitudinal, multicenter From the Division of Cardiology, Department of Medicine, University of California, Irvine, California; Department of Biostatistics, University of Washington, Seattle, Washington; Division of Cardiology, Department of Medicine, Wake Forest University School of Medicine, Winston-Salem, North Carolina; Departments of Medicine and Epidemiology, The Johns Hopkins Medical Institutions, Baltimore, Maryland; Division of Cardiology, University of California, Davis, California; Division of Cardiology, Albany Medical College, Albany, New York; and Division of Cardiology, Department of Medicine, Georgetown University Medical Center, Washington, DC. This study was supported by Contracts Nos. NO1-85079 to HC-85086 from the National Heart, Lung, and Blood Institute, Bethesda, Maryland. Manuscript received July 31, 2000; revised manuscript received and accepted December 5, 2000. Address for reprints: Julius M. Gardin, MD, St. John Guild Distinguished Chair, Chief, Division of Cardiology, St. John Hospital & Medical Center, 22151 Moross Road, Detroit, Michigan 48236. E-mail: [email protected]. ©2001 by Excerpta Medica, Inc. All rights reserved. The American Journal of Cardiology Vol. 87 May 1, 2001
METHODS
Subjects: CHS is a prospective, population-based study of 5,888 participants (2,495 men and 3,393 women) ⱖ65 years old. Of this cohort, 4,926 participants were white, 917 were black, and 55 were classified as “other nonwhite” at baseline. The cohort included 5,201 men and women (244 blacks) recruited at baseline (“initial cohort”) in whom 6 to 7 years of follow-up were available and an additional 687 participants (including 673 blacks) who were recruited subsequently (“new cohort”). Ages ranged from 65 to 100 years (mean ⫾ SD 73.3 ⫾ 5.8 years in men and 0002-9149/01/$–see front matter PII S0002-9149(01)01460-6
1051
72.5 ⫾ 5.5 years in women). CHS participants were recruited and examined at 4 field centers: Forsyth County, North Carolina; Sacramento County, California; Allegheny County, Pennsylvania; and Washington County, Maryland. The overall design, objectives, recruitment strategy, and cohort characteristics of CHS have been previously published.13,14 Definition of prevalent and incident disease: In the present study, all analyses excluded subjects (n ⫽ 2,108) with prevalent disease at entry into the study. Prevalent disease was defined as a self-reported history of CHD, stroke, CHF, atrial fibrillation (either by electrocardiogram or self-report), pacemaker implant, use of digoxin, or severe aortic and/or mitral regurgitation by Doppler echocardiography.15 CHD was defined as confirmed myocardial infarction or angina pectoris, or self-reported history of coronary artery bypass surgery or coronary artery angioplasty. CHF (or stroke) was defined as a “yes” response to the question: “Has a doctor ever told you that you had congestive heart failure (or stroke)?” and/or confirmation of the history of CHF (or stroke) by examination or medical records review. CHD, CHF, or stroke events not reported by the participant, but noted in the medical records, were also categorized as prevalent disease. Outcomes of interest included incident CHD, incident CHF, incident stroke, and all-cause mortality. All incident events were evaluated by a panel of investigators to determine their classification (i.e., stroke, CHF, and so forth) or cause of death. Details of the adjudication process have been previously described.16 For the initial cohort, 6 to 7 years of follow-up data were available, whereas for the new cohort, only 1 to 2 years of follow-up were available after echocardiography. Echocardiographic performance and reading protocol: Echocardiography was performed in 1988 to 1989
during a baseline CHS examination on the “initial cohort” that included interviews related to medical history, physical activity, personal habits, cognitive function and dietary intake; anthropometry; recumbent, sitting, and standing blood pressure; resting and ambulatory electrocardiography; spirometry; carotid ultrasound; and laboratory studies.13 For the “new cohort,” echocardiography was performed in 1994 to 1995, 2 years after their entry into the study. The CHS echocardiographic protocol and quality control measures have also been described.17 Echocardiograms were recorded onto superVHS tape using a standardized protocol. Measurements were made at a core reading center from digitized images using off-line image analysis systems equipped with customized computer algorithms. Echocardiographic measurements: This study focuses on 2-dimensional M-mode measurements of LV mass and its 3 components measured at end-diastole— ventricular septal thickness (VSTd), LV (internal) dimension (LVIDd), and LV posterior wall thickness (PWTd)—and on the left atrium. M-mode measurements were made following recommendations of the American Society of Echocardiography.18 LV 1052 THE AMERICAN JOURNAL OF CARDIOLOGY姞
VOL. 87
TABLE 1 Descriptors of Left Ventricular (LV) Geometry Categories LV Geometry Category Normal Concentric remodeling Eccentric LV hypertrophy Concentric LV hypertrophy
LV Relative Wall Thickness
LV Mass Index
Normal Abnormal Normal Abnormal
Normal Normal Abnormal Abnormal
mass was derived as described by Devereux et al19: LV (grams) ⫽ 0.80 ⫻ 1.04 [(VSTd ⫹ LVIDd ⫹ PWTd)3 ⫺ (LVIDd)3] ⫹ 0.6, where thickness and dimension measurements are expressed in centimeters. LV relative wall thickness was calculated as twice the ratio of PWTd to LVIDd.4,20 Echocardiograms of adequate quality for performing M-mode measurements of the left ventricle were available in 1,488 men and 2,322 women. As previously described, ⱖ1 LV measurements could not be made in 34% of the elderly CHS cohort, with age being the strongest correlate of missing data.21 After excluding participants with prevalent disease at baseline, 884 men and 1,622 women remained and formed the basis for this study. Explanatory echocardiographic variables of interest included LV mass, LV internal dimensions in diastole and systole, ventricular septal thickness in diastole, LV posterior wall thickness in diastole, left atrial dimension, and mitral annular calcium (yes/no). In addition, a categorical variable for LV geometry and/or hypertrophy was defined, based on abnormal categories of LV relative wall thickness,4,20 and LV mass index (ratio of observed LV mass to predicted LV mass based on height and gender). Predicted LV mass was based on the results of a regression of observed LV mass (log transformed to increase normality) on height and gender.22 Abnormal LV relative wall thickness was defined to be any value more than the 95th percentile (⬎0.48); abnormal LV mass index was defined to be any value more than the 95th percentile (⬎1.635 predicted). LV geometry categories are defined in Table 1. Statistical analysis: Because baseline echocardiographic measurements were missing for many participants without prevalent disease who would otherwise have been included in the model (n ⫽ 1,274 missing values for most echocardiographic variables, but only 216 for left atrial dimension), incidence rates (per 1,000 person–years at risk) were compared between those with and without echocardiographic measurements. For all continuous predictor variables, age and cohort-adjusted means stratified by gender and incident disease status were calculated for each outcome of interest. Analysis of covariance was used to investigate which differences were significant. Cox regression was used to model the time to incident disease as a function of each echocardiographic variable individually. Time-to-event plots (stratified by the levels of each predictor of interest) were examined, and there was no evidence to indicate MAY 1, 2001
TABLE 2 Age and Cohort Adjusted Means of M-Mode Echocardiographic Variables by Incident Disease Status (excluding participants with prevalent disease at baseline) Incident CHD Echocardiographic Measure LV mass (g) (n ⫽ 2,506) Women Men LV dimension (diastole, mm) (n ⫽ 2,506) Women Men LV dimension (systole, mm) (n ⫽ 2,492) Women Men Ventricular septal thickness (diastole, mm) (n ⫽ 2,506) Women Men LV posterior wall thickness (diastole, mm) (n ⫽ 2,506) Women Men Left atrial dimension (mm) (n ⫽ 3,564) Women Men
Incident CHF
Incident Stroke
Mortality (All-cause)
⫹
o
⫹
o
⫹
o
⫹
o
136.5* 180.4†
130.0 167.9
152.4‡ 198.7‡
129.3 167.6
140.8* 182.8
130.5 169.6
135.7 183.1†
130.6 168.1
47.4 51.7
47.0 51.6
49.0 53.5†
46.8 51.3
48.4* 50.9
46.8 51.5
46.7 51.9
47.0 51.3
27.4 31.4
26.7 31.6
29.5‡ 33.5‡
26.5 30.4
28.0* 30.8
26.6 30.7
26.7 31.9*
26.6 30.4
8.7 9.7†
8.4 9.2
9.2‡ 10.1‡
8.5 9.2
8.6 9.9*
8.5 9.2
8.7 9.7†
8.5 9.2
8.4* 9.1†
8.1 8.8
8.6† 9.4†
8.1 8.8
8.3 9.4†
8.2 8.8
8.3 9.2†
8.2 8.8
37.0 39.4
38.8† 41.4†
37.0 39.3
37.7 39.9
‡
38.2 39.9
37.1 39.5
37.1 39.7
37.2 39.5
*p ⬍0.05; †p ⬍0.01; ‡p ⬍0.001 based on analysis of variance for equality of echocardiographic measurement means between disease status groups with gender.
that the proportional hazards assumption was inappropriate. Both unadjusted models and models including other cardiovascular risk factors are presented. Risk factors included age, gender, high-density lipoprotein cholesterol, weight, height, smoking status (current, former, never), hypertension status, and diabetes (normal, glucose intolerant, or diabetic) based on the new American Diabetes Association criteria. The echocardiographic variables were entered first as continuous variables, and then divided into gender-specific quartiles to capture possible nonlinearities in the relations. All models were checked for interactions between gender and echocardiographic measurements of interest. SPSS software (version 7.5, SPSS Inc., Chicago, Illinois) was used for all statistical analyses.
RESULTS
Missing echocardiographic measurements: CHS participants with missing echocardiographic LV measurements demonstrated a significantly higher rate of all-cause mortality (33.9 vs 16.5 per 1,000 person– years at risk) than those in whom echocardiographic measurements could be performed. Of importance, however, there were no differences in the rate of incident CHD, CHF, or stroke between the 2 groups.
Echocardiographic measurements per incident disease and gender (Table 2): When age-adjusted means
of M-mode echocardiographic variables were considered by gender, all measurements, in all incident disease status subgroups, were larger in men than women. As previously shown for LV mass, this difference in CHS was only partially explained by the larger body size of the men.21 When incident disease
status subgroups were considered, LV mass was significantly greater in men and women with, versus without, incident CHD and incident CHF; in women with, versus women without, stroke; and in men with, versus men without, death from all causes. There was a 23- to 31-g mean difference in age-adjusted mass in women and men with, versus without incident CHF. LV diastolic and systolic dimensions were greater in men and women with, versus without incident CHF (by 2.2 to 3.1 mm, on average), and greater in women with, versus without, incident stroke. LV systolic dimension was greater in men who died from all causes. Ventricular septal thickness was greater in men and women with, versus without, incident CHF and in men with, versus without, incident CHD, stroke, and allcause mortality. LV posterior wall thickness was greater in men and women with, versus without, incident CHD and CHF, and greater in men with, versus without, incident stroke and all-cause mortality. Finally, left atrial dimension was larger in men and women with, versus without, incident CHF. Echocardiographic measurement quartiles and allcause mortality: Unadjusted LV mass, ventricular sep-
tal thickness, posterior wall thickness, and left atrial dimension all showed a significant positive relation with all-cause mortality (range of hazards ratios for the highest vs lowest quartiles 1.15 to 1.80). Presence of mitral annular calcium also conferred increased risk for all-cause mortality (hazards ratio 1.71, p ⬍0.001). However, after adjustment for traditional risk factors, there remained significant relations only between allcause mortality and 2 echocardiographic variables:
CORONARY ARTERY DISEASE/ECHOCARDIOGRAPHIC PREDICTORS OF CARDIOVASCULAR EVENTS
1053
TABLE 3 Associations of Echocardiographic Measures and Incident Coronary Heart Disease (CHD) Incident CHD Echocardiographic Measure LV mass (g) Quartiles 1 2 3 4 LV dimension (diastole, mm) Quartiles 1 2 3 4 LV dimension (systole, mm) Quartiles 1 2 3 4 Ventricular septal thickness (diastole, mm) Quartiles 1 2 3 4 LV posterior wall thickness (diastole, mm) Quartiles 1 2 3 4 Left atrial dimension (mm) Quartiles 1 2 3 4 Mitral annular calcium No (n ⫽ 3,429) Yes (n ⫽ 244) LV anatomy category Height/gender adjustment Normal (n ⫽ 2,318) Concentric remodeling (n ⫽ 110) Eccentric LV hypertrophy (n ⫽ 55) Concentric LV hypertrophy (n ⫽ 14)
No. of Incident CHD Cases
Rate per 1,000 Person–Years
Unadjusted Hazards Ratio (95% CI)
Hazards Ratio Adjusted for Risk Factors (95% CI)
57 75 71 89
16.6 21.9 20.6 26.6
p ⬍0.001 1.00 1.34 (0.95–1.89) 1.25 (0.89–1.78) 1.63 (1.17–2.27)
p ⫽ 0.0053 1.00 1.33 (0.94–1.89) 1.22 (0.85–1.74) 1.43 (0.99–2.05)
66 72 76 78
19.9 21.6 22.0 21.7
p ⫽ 0.010 1.00 1.08 (0.78–1.51) 1.10 (0.79–1.53) 1.08 (0.78–1.50)
p ⫽ 0.9672 1.00 1.11 (0.79–1.55) 1.12 (0.80–1.57) 1.04 (0.74–1.47)
71 69 63 89
21.2 20.1 18.3 26.2
p ⫽ 0.001 1.00 0.95 (0.68–1.32) 0.86 (0.62–1.21) 1.23 (0.90–1.69)
p ⫽ 0.0775 1.00 0.89 (0.64–1.25) 0.89 (0.63–1.25) 1.24 (0.90–1.71)
65 68 76 83
18.2 19.7 22.6 25.2
p ⬍0.001 1.00 1.08 (0.77–1.52) 1.25 (0.90–1.74) 1.40 (1.01–1.93)
p ⫽ 0.0043 1.00 1.08 (0.76–1.52) 1.19 (0.85–1.67) 1.19 (0.85–1.67)
64 59 86 83
17.4 17.3 24.8 26.4
p ⬍0.001 1.00 1.00 (0.70–1.42) 1.44 (0.04–1.99) 1.54 (1.11–2.13)
p ⫽ 0.0021 1.00 0.94 (0.66–1.34) 1.32 (0.95–1.84) 1.33 (0.94–1.86)
93 115 115 113
18.3 23.6 23.8 24.0
408 47
21.7 35.4
p ⬍0.001 1.00 1.29 (0.98–1.70) 1.31 (0.99–1.72) 1.32 (1.00–1.74) p ⫽ 0.0015 1.00 1.63 (1.21–2.21)
p ⫽ 0.2402 1.00 1.24 (0.94–1.63) 1.21 (0.92–1.61) 1.21 (0.91–1.61) p ⫽ 0.0290 1.00 1.41 (1.04–1.93)
258 18 13 2
33.7 47.9 36.4 26.4
p ⫽ 0.0029 1.00 1.69 (1.05–2.73) 2.39 (1.37–4.18) 1.88 (0.47–7.54)
p ⫽ 0.0246 1.00 1.57 (0.97–2.54) 2.05 (1.16–3.62) 1.61 (0.40–6.54)
The p value for each echocardiographic variable represents the overall significance of the continuous variable in the model. CI ⫽ confidence interval.
LV mass (hazards ratio 1.21, p ⬍0.0122) and posterior wall thickness (hazards ratio 1.33, p ⬍0.0175). Of interest, there were positive relations of eccentric LV hypertrophy and concentric remodeling with all-cause mortality. There were no gender differences in associations reported in the following between echocardiographic variables and incident all-cause mortality, CHD, stroke, or CHF. Echocardiographic measurements and incident coronary heart disease (Table 3): Before adjustment for
traditional risk factors, all echocardiographic variables conferred an increased risk of incident CHD. Even after adjustment, the highest (versus lowest) quartiles of LV mass, septal thickness, posterior wall thickness, and the presence of mitral annular calcium conferred 1054 THE AMERICAN JOURNAL OF CARDIOLOGY姞
VOL. 87
increased risks for incident CHD (range of hazards ratios 1.2 to 1.4). Furthermore, the eccentric and concentric LV hypertrophy and concentric remodeling patterns conferred increased hazards ratios (range 1.6 to 2.0) for incident CHD, after adjustment for risk factors. Figure 1 illustrates the percent of subjects who were free of incident CHD, stratified by genderspecific quartiles of LV mass. Those in the lowest LV mass quartile had the longest CHD-free survival, with approximately 90% of participants free of incident CHD by the end of the study period— compared with 80% free of CHD in the highest quartile. The middle 2 quartiles were intermediate in terms of CHD-free survival. None of the curves were signifiMAY 1, 2001
DISCUSSION
FIGURE 1. Kaplan-Meier curves for incident CHD by LV mass gender-specific quartiles. The graph displays cumulative incident CHD-free survival (in days) in the 4 gender-specific LV mass quartiles. The vertical axis indicates the percent of the original study participants (who did not have a history of CHD) who remained free of incident CHD by each time point. The numbers 1 to 4 indicate the lowest to highest gender-specific quartiles of LV mass.
cantly different from one another (although there was a linear trend indicated by the overall p value of 0.0053). Echocardiographic measurements and incident stroke: Before adjustment for traditional risk factors,
LV mass, septal and posterior wall thickness, left atrial dimension, and the presence of mitral annular calcium conferred a significantly increased risk of incident stroke. However, after adjustment for risk factors, only echocardiographic LV mass conferred an increased risk (hazards ratio 1.52, highest vs lowest quartile) for incident stroke. None of the LV geometry categories conferred an increased risk for incident stroke, possibly because of the small number of cases in the 2 LV hypertrophy groups. Echocardiographic measurements and incident congestive heart failure (Table 4): All echocardiographic
variables (unadjusted for risk factors) were positively associated with incident CHF (range of unadjusted hazards ratios 1.9 to 4.1). Significant hazards ratios (e.g., 3.4 for LV mass) persisted for all of these echocardiographic variables, except for mitral annular calcium, after adjustment for risk factors. Both concentric and eccentric hypertrophy were associated with increased hazards ratios for incident CHF (3.3 and 3.0, respectively, after risk factor adjustment). Figure 2 illustrates the percent of the cohort who were free of incident CHF stratified by gender-specific quartiles of LV mass. The lowest 2 quartiles had equivalent curves, with approximately 97% free of incident CHF by the end of the study. In contrast, both the third, and particularly the highest quartile of mass, exhibited significantly worse curves than the 2 lowest quartiles, with 92% and 86% free of incident CHF, respectively, by the end of the study.
Present study findings: The present study demonstrates, in the elderly event-free CHS cohort, the predictive value of 2-dimensional M-mode echocardiographic measurements. Specifically, relations remained significant, even after adjustment for traditional risk factors, between: (1) incident CHF and LV mass, internal diastolic and systolic dimensions, septal and posterior wall thickness, concentric and eccentric LV hypertrophy, and left atrial dimension; (2) incident CHD and LV mass, septal and posterior wall thickness, eccentric hypertrophy and mitral annular calcium, and (3) all-cause mortality and LV mass, posterior wall thickness, and non-normal ventricular geometry category. These data extend previous observations, because unlike the Framingham Heart Study, CHS is a multicenter study, specifically of the elderly, and includes a relatively large cohort of blacks, as well as whites. Ventricular mass and incident events: In CHS, LV mass was linearly associated with all-cause mortality even after adjustment for traditional risk factors. More striking, the hazards ratios in CHS for incident CHF for the third and fourth quartiles of adjusted mass were 2.0 and 3.4, respectively (compared with the lowest quartile). In contrast, the hazards ratios for incident CHD in CHS were less than relative risks (7.8 in men and 3.4 in women) for CHD events conferred by the highest versus the lowest quartile of risk factor-adjusted mass reported in a 4-year follow-up in the original cohort of the Framingham study (mean age 69 years).2 The explanation for the different magnitudes of hazards ratios between the 2 studies is unclear. Ventricular geometry and incident events: Increased LV internal dimensions in the CHS cohort were positively associated with incident CHF, similar to previous study from Framingham investigators.6,7 In addition, in CHS, ventricular septal and posterior wall thickness were positively associated with incident CHF. Eccentric LV hypertrophy and concentric remodeling, but not concentric hypertrophy, conferred increased hazards ratios for incident CHD, even after adjustment for traditional risk factors. In contrast, Koren et al4 reported that in patients with uncomplicated essential hypertension, concentric and eccentric hypertrophy were importantly related to the incidence of CVD events (p ⬍0.03) and mortality (p ⬍0.001). LV mass and geometry stratified risk independently of, and more strongly than, blood pressure or other potentially reversible risk factors. Similarly, among Framingham subjects ⱖ40 years old who were free of clinical CVD, those with concentric hypertrophy had the worst prognosis, followed by those with eccentric hypertrophy, concentric remodeling, and normal geometry.23 Concentric hypertrophy may not have reached statistical significance as a predictor of incident CHD in CHS because of the small number (n ⫽ 2) of incident CHD cases. Ventricular geometry, mass, and stroke: In CHS, after adjustment for traditional risk factors, LV mass was associated with a modestly increased hazards ratio (1.52 in the highest quartile) for stroke. No LV
CORONARY ARTERY DISEASE/ECHOCARDIOGRAPHIC PREDICTORS OF CARDIOVASCULAR EVENTS
1055
TABLE 4 Associations of Echocardiographic Measurements and Incident Congestive Heart Failure (CHF) Incident CHF Echocardiographic Measure LV mass (g) Quartiles 1 2 3 4 LV dimension (diastole, mm) Quartiles 1 2 3 4 LV dimension (systole, mm) Quartiles 1 2 3 4 Ventricular septal thickness (diastole, mm) Quartiles 1 2 3 4 LV posterior wall thickness (diastole, mm) Quartiles 1 2 3 4 Left atrial dimension (mm) Quartiles 1 2 3 4 Mitral annular calcium No (n ⫽ 3,428) Yes (n ⫽ 244) LV anatomy category Height/gender adjustment Normal (n ⫽ 2,320) Concentric remodeling (n ⫽ 11) Eccentric LV hypertrophy (n ⫽ 52) Concentric LV hypertrophy (n ⫽ 13)
No. of Incident CHF Cases
Rate per 1000 Person–Years
Unadjusted Hazards Ratio (95% CI)
Hazards Ratio Adjusted for Risk Factors (95% CI)
19 19 41 73
5.3 5.3 11.5 21.6
p ⬍0.001 1.00 1.00 (0.53–1.88) 2.18 (1.27–3.76) 4.11 (2.48–6.81)
p ⬍0.001 1.00 1.00 (0.53–1.89) 2.01 (1.16–3.51) 3.36 (1.96–5.74)
26 33 36 57
7.6 9.5 10.1 15.7
p ⬍0.001 1.00 1.25 (0.75–2.09) 1.32 (0.80–2.19) 2.05 (1.29–3.27)
p ⬍0.001 1.00 1.33 (0.79–2.23) 1.44 (0.86–2.41) 2.05 (1.26–3.33)
28 24 30 70
7.9 6.7 8.5 20.4
p ⬍0.001 1.00 0.84 (0.49–1.45) 1.07 (0.64–1.80) 2.57 (1.66–3.99)
p ⬍0.001 1.00 0.80 (0.46–1.39) 1.20 (0.71–2.02) 2.69 (1.71–4.23)
22 29 46 55
6.0 8.1 13.3 16.2
p ⬍0.001 1.00 1.37 (0.79–2.38) 2.25 (1.36–3.74) 2.76 (1.69–4.53)
p ⬍0.001 1.00 1.22 (0.70–2.12) 1.87 (1.12–3.13) 1.87 (1.13–3.11)
24 28 42 58
6.3 8.0 11.7 18.0
p ⬍0.001 1.00 1.28 (0.74–2.20) 1.88 (1.14–3.11) 2.94 (1.83–4.73)
p ⬍0.001 1.00 1.10 (0.63–1.90) 1.47 (0.88–2.45) 2.00 (1.22–3.27)
40 52 60 80
7.6 10.3 12.0 16.7
216 29
11.2 21.2
p ⬍0.001 1.00 1.35 (0.90–2.04) 1.57 (1.06–2.35) 2.21 (1.51–3.23) p ⫽ 0.0012 1.00 1.89 (1.29–2.79)
p ⬍0.001 1.00 1.28 (0.84–1.94) 1.49 (0.99–2.24) 1.90 (1.28–2.82) p ⫽ 0.1454 1.00 1.34 (0.90–2.00)
129 9 11 3
9.8 15.9 40.6 57.0
p ⬍0.001 1.00 1.66 (0.84–3.26) 4.27 (2.31–7.91) 6.26 (1.99–19.67)
p ⫽ 0.0018 1.00 1.40 (0.71–2.76) 2.95 (1.56–5.57) 3.32 (1.04–10.60)
The p value for each echocardiographic variable represents the overall significance of the continuous variable in the model. Abbreviation as in Table 3.
geometric pattern conferred an increased adjusted hazards ratios for incident stroke. However, geometric subgroup analyses were limited due to the small numbers of events–for example, there were only 2 incident stroke cases in the concentric hypertrophy subgroup. Recent studies have documented the association of carotid plaque and stenosis, and increased intimalmedial thickness, with higher LV mass.24,25 Left atrial dimension and incident events: In CHS, left atrial dimension was predictive of all-cause mortality, incident stroke, and CHD; however, these associations became nonsignificant after adjustment for traditional risk factors. In contrast, left atrial dimension, after adjustment, remained predictive of incident CHF. These results differ from both the Framingham 1056 THE AMERICAN JOURNAL OF CARDIOLOGY姞
VOL. 87
Heart Study8 and the Stroke Prevention in Atrial Fibrillation trial,9,10 in which left atrial dimension was related to strokes and death in the former, and to thromboembolic events (e.g., strokes and transient ischemic attacks) in the latter. Different findings in these 3 studies may relate to differences in subject ages and selection criteria. Mitral annular calcium and incident events: In CHS, mitral annular calcium conferred an increased unadjusted hazards ratio for all-cause mortality, incident CHD, stroke, and CHF, but remained predictive only of incident CHD after adjustment for risk factors. In contrast, in the Framingham study,11 mitral annular calcium conferred a greater risk for incident stroke than did atrial fibrillation. There has been speculation MAY 1, 2001
FIGURE 2. Kaplan-Meier curves for incident CHF by LV mass gender-specific quartiles. The graph displays cumulative incident CHF-free survival (in days) in the 4 gender-specific LV mass quartiles. The format is similar to that for Figure 1.
regarding the pathophysiologic explanation for the relation of mitral annular calcium and various incident CVD events. Similar risk factors may be involved in genesis of both mitral annular calcium and CHD. This hypothesis is strengthened by the association found in CHS between coronary risk factors and aortic valve sclerosis.26 In summary, this study suggests that in an elderly population, selected M-mode echocardiographic measurements are subclinical markers that confer independent prognostic information for incident CVD events, particularly for CHD and CHF.13 1. Casale PN, Devereux RB, Milner M, Zullo G, Pickering TG, Laragh JH. Value
of echocardiographic measurement of left ventricular geometry mass in predicting cardiovascular morbid events in hypertensive men. Ann Intern Med 1986; 105:173–178. 2. Levy D, Garrison RJ, Savage DD, Kannell WB, Castelli WP. Left ventricular geometry, mass and incidence of coronary heart disease in an elderly cohort. The Framingham Heart Study. Ann Intern Med 1989;110:101–107. 3. Levy D, Garrison RJ, Savage DD, Kannell WB, Castelli WP, Laragh JH. Prognostic implications of echocardiographically determined left ventricular geometry and mass in the Framingham Heart Study. N Engl J Med 1990;322:1561– 1566. 4. Koren MJ, Devereux RB, Casale PN, Savage DD. Relation of left ventricular geometry mass and geometry to morbidity and mortality in uncomplicated essential hypertension. Ann Intern Med 1991;114:345–352. 5. Ghali JK, Liao Y, Simmons B, Castaner A, Cao, G, Cooper RS. The prognostic role of left ventricular geometry hypertrophy in patients with or without coronary artery disease. Ann Intern Med 1992;117:831– 836. 6. Lauer MS, Evans JC, Levy D. Prognostic implications of subclinical left
ventricular dilatation and systolic dysfunction in men free from overt cardiovascular disease (the Framingham Heart Study). Am J Cardiol 1992;70:1180 –1184. 7. Vasan RS, Larson MG, Benjamin EJ, Evans JC, Levy, D. Left ventricular dilatation and the risk of congestive heart failure in people without myocardial infarction. N Engl J Med 1997;336:1350 –1355. 8. Benjamin EJ, D’Agostino RB, Belanger AJ, Wolf PA, Levy D. Left atrial size and the risk of stroke and death: the Framingham Heart Study. Circulation 1995;92:835– 841. 9. The Stroke Prevention in Atrial Fibrillation Investigators. Predictors of thromboembolism in atrial fibrillation: I. Clinical features of patients at risk. Ann Intern Med 1992;116:1–5. 10. The Stroke Prevention in Atrial Fibrillation Investigators. Predictors of thromboembolism in atrial fibrillation: II. echocardiographic features of patients at risk. Ann Intern Med 1992;116:6 –12. 11. Benjamin EJ, Plehn JF, D’Agostino RB, Belanger AJ, Comai K, Fuller, DL, Wolf, PA, Levy D. Mitral annular calcification and the risk of stroke in an elderly cohort. N Engl J Med 1992;327:374 –379. 12. Devereux RB, Alderman MH. Role of preclinical cardiovascular disease in the evolution from risk factors exposure to development of morbid events. Circulation 1993;88:1444 –1455. 13. Fried L, Borhani N, Enright P, Furberg CD, Gardin JM, Kronmal RA, Kuller LH, Manolio TA, Mittelmark MB, Newman A. Cardiovascular Health Study. Design and rationale. The CHS Collaborative Research Group. Ann Epidemiol 1991;1:263–276. 14. Tell G, Fried L, Hermanson B, Manolio TA, Newman AB, Borhani N, for the Cardiovascular Health Study Collaborative Research Group. Recruitment of adults 65 years and older as participants in the Cardiovascular Health Study. Ann Epidemiol 1993;3:358 –366. 15. Psaty B, Kuller LH, Bild D, Burke GL, Kittner SJ, Mittlemark M, Price TR, Rautaharju PM, Robbins J. Methods of assessing prevalent cardiovascular disease in the Cardiovascular Health Study. Ann Epidemiol 1995;5:270 –277. 16. Ives DG, Fitzpatrick AL, Bild DE, Psaty BM, Kuller KH, Crowley PM, Cruise RG, Theroux S. Surveillance and ascertainment of cardiovascular events: the Cardiovascular Health Study. Ann Epidemiol 1995;5:278 –285. 17. Gardin JM, Wong ND, Bommer W, Klopfenstein HS, Smith VE, Tabatznik B, Siscovick D, Lobodzinski S, Anton-Culver H, Manolio TA. Echocardiographic design of a multi-center investigation of free-living elderly subjects: the Cardiovascular Health Study. J Am Soc Echocardiogr 1992;5:63–72. 18. Sahn DJ, DeMaria A, Kisslo J, Weyman A, the Committee on M-mode Standardization of the American Society of Echocardiography. Recommendations regarding quantitation in M-mode echcoardiography: results of a survey of echocardiographic methods. Circulation 1978;58:1072–1083. 19. Devereux RB, Alonso DR, Lutas EM, Gottlieb GJ, Campo E, Sachs I, Reichek N. Echocardiographic assessment of left ventricular geometry hypertrophy: comparison with necropsy findings. Am J Cardiol 1986;57:450 – 458. 20. Ganau A, Devereux RB, Roman MJ, de Simone G, Pickering TG, Saba PS, Vargiu P, Simongini I, Laragh JH. Patterns of left ventricular geometry hypertrophy and geometric remodelling in essential hypertension. J Am Coll Cardiol 1992;19:1550 –1558. 21. Gardin JM, Siscovick D, Anton-Culver H, Lynch JC, Smith VE, Klopfenstein HS, Bommer WJ, Fried L, O’Leary D, Manolio TA. Sex, age and disease affect echocardiographic left ventricular geometry, mass, and systolic function in the free-living elderly: the Cardiovascular Health Study (CHS). Circulation 1995; 91:1739 –1748. 22. Lauer MS, Larson MG, Levy D. Gender-specific reference M-mode values in adults: population-derived values with consideration of the impact of height. J Am Coll Cardiol 1995;26:1039 –1046. 23. Krumholz HM, Larson M, Levy D. Prognosis of left ventricular geometric patterns in the Framingham Heart Study. J Am Coll Cardiol 1995;25:879 – 884. 24. Roman MJ, Pickering TG, Schwartz JE, Pini R, Devereux RB. Association of carotid atherosclerosis and left ventricular hypertrophy. J Am Coll Cardiol 1995;25:83–90. 25. Kronmal RA, Smith VE, O’Leary DH, Polak JF, Gardin, Manolio TA. Carotid artery measures are strongly associated with left ventricular mass in older adults: a report from the Cardiovascular Health Study. Am J Cardiol 1996;77: 628 – 633. 26. Stewart BF, Siscovick D, Lind BK, Gardin JM, Gottdiener JS, Smith VE, Kitzman DW, Otto CM. Clinical factors associated with calcific aortic valve disease. J Am Coll Cardiol 1997;29:630 – 634.
CORONARY ARTERY DISEASE/ECHOCARDIOGRAPHIC PREDICTORS OF CARDIOVASCULAR EVENTS
1057
stroke. After adjustment for anthropometric and traditional CVD risk factors, LV mass was significantly related to incident CHD, CHF, and stroke. The highest quartile of LV mass conferred a hazards ratio of 3.36, compared with the lowest quartile, for incident CHF. Furthermore, incident CHF-free survival was significantly lower for participants with LV mass in the highest versus the 2 lowest quartiles (86% vs 97%, respectively, at 2,500 days). Eccentric and concentric LV hypertrophy, respectively, conferred adjusted hazards ratios, compared with normal LV geometry, of 2.05 and 1.61 for incident CHD, and 2.95 and 3.32 for incident CHF. Thus, in an elderly biracial population, selected 2-dimensional Mmode echocardiographic measurements were important markers of subclinical disease and conferred independent prognostic information for incident CVD events, especially CHF and CHD. 䊚2001 by Excerpta Medica, Inc. (Am J Cardiol 2001;87:1051–1057)
revious studies have identified a number of transthoracic M-mode echocardiographic variables P with independent prognostic value for cardiovascular
cohort study of elderly men and women sponsored by the National Heart, Lung, and Blood Institute. This study focuses on a 6- to 7-year follow-up period, examining the relation of M-mode echocardiographic variables, including left ventricular (LV) mass and its component variables, LV geometric subtypes (including concentric and eccentric hypertrophy and concentric remodeling), left atrial dimension, and mitral annular calcium, to incident coronary heart disease (CHD), congestive heart failure (CHF), stroke, and (cardiac and overall) mortality.13,14
disease (CVD) events and overall mortality. However, these studies have generally not focused on an elderly population-based cohort.1–12 The Cardiovascular Health Study (CHS) is a longitudinal, multicenter From the Division of Cardiology, Department of Medicine, University of California, Irvine, California; Department of Biostatistics, University of Washington, Seattle, Washington; Division of Cardiology, Department of Medicine, Wake Forest University School of Medicine, Winston-Salem, North Carolina; Departments of Medicine and Epidemiology, The Johns Hopkins Medical Institutions, Baltimore, Maryland; Division of Cardiology, University of California, Davis, California; Division of Cardiology, Albany Medical College, Albany, New York; and Division of Cardiology, Department of Medicine, Georgetown University Medical Center, Washington, DC. This study was supported by Contracts Nos. NO1-85079 to HC-85086 from the National Heart, Lung, and Blood Institute, Bethesda, Maryland. Manuscript received July 31, 2000; revised manuscript received and accepted December 5, 2000. Address for reprints: Julius M. Gardin, MD, St. John Guild Distinguished Chair, Chief, Division of Cardiology, St. John Hospital & Medical Center, 22151 Moross Road, Detroit, Michigan 48236. E-mail: [email protected]. ©2001 by Excerpta Medica, Inc. All rights reserved. The American Journal of Cardiology Vol. 87 May 1, 2001
METHODS
Subjects: CHS is a prospective, population-based study of 5,888 participants (2,495 men and 3,393 women) ⱖ65 years old. Of this cohort, 4,926 participants were white, 917 were black, and 55 were classified as “other nonwhite” at baseline. The cohort included 5,201 men and women (244 blacks) recruited at baseline (“initial cohort”) in whom 6 to 7 years of follow-up were available and an additional 687 participants (including 673 blacks) who were recruited subsequently (“new cohort”). Ages ranged from 65 to 100 years (mean ⫾ SD 73.3 ⫾ 5.8 years in men and 0002-9149/01/$–see front matter PII S0002-9149(01)01460-6
1051
72.5 ⫾ 5.5 years in women). CHS participants were recruited and examined at 4 field centers: Forsyth County, North Carolina; Sacramento County, California; Allegheny County, Pennsylvania; and Washington County, Maryland. The overall design, objectives, recruitment strategy, and cohort characteristics of CHS have been previously published.13,14 Definition of prevalent and incident disease: In the present study, all analyses excluded subjects (n ⫽ 2,108) with prevalent disease at entry into the study. Prevalent disease was defined as a self-reported history of CHD, stroke, CHF, atrial fibrillation (either by electrocardiogram or self-report), pacemaker implant, use of digoxin, or severe aortic and/or mitral regurgitation by Doppler echocardiography.15 CHD was defined as confirmed myocardial infarction or angina pectoris, or self-reported history of coronary artery bypass surgery or coronary artery angioplasty. CHF (or stroke) was defined as a “yes” response to the question: “Has a doctor ever told you that you had congestive heart failure (or stroke)?” and/or confirmation of the history of CHF (or stroke) by examination or medical records review. CHD, CHF, or stroke events not reported by the participant, but noted in the medical records, were also categorized as prevalent disease. Outcomes of interest included incident CHD, incident CHF, incident stroke, and all-cause mortality. All incident events were evaluated by a panel of investigators to determine their classification (i.e., stroke, CHF, and so forth) or cause of death. Details of the adjudication process have been previously described.16 For the initial cohort, 6 to 7 years of follow-up data were available, whereas for the new cohort, only 1 to 2 years of follow-up were available after echocardiography. Echocardiographic performance and reading protocol: Echocardiography was performed in 1988 to 1989
during a baseline CHS examination on the “initial cohort” that included interviews related to medical history, physical activity, personal habits, cognitive function and dietary intake; anthropometry; recumbent, sitting, and standing blood pressure; resting and ambulatory electrocardiography; spirometry; carotid ultrasound; and laboratory studies.13 For the “new cohort,” echocardiography was performed in 1994 to 1995, 2 years after their entry into the study. The CHS echocardiographic protocol and quality control measures have also been described.17 Echocardiograms were recorded onto superVHS tape using a standardized protocol. Measurements were made at a core reading center from digitized images using off-line image analysis systems equipped with customized computer algorithms. Echocardiographic measurements: This study focuses on 2-dimensional M-mode measurements of LV mass and its 3 components measured at end-diastole— ventricular septal thickness (VSTd), LV (internal) dimension (LVIDd), and LV posterior wall thickness (PWTd)—and on the left atrium. M-mode measurements were made following recommendations of the American Society of Echocardiography.18 LV 1052 THE AMERICAN JOURNAL OF CARDIOLOGY姞
VOL. 87
TABLE 1 Descriptors of Left Ventricular (LV) Geometry Categories LV Geometry Category Normal Concentric remodeling Eccentric LV hypertrophy Concentric LV hypertrophy
LV Relative Wall Thickness
LV Mass Index
Normal Abnormal Normal Abnormal
Normal Normal Abnormal Abnormal
mass was derived as described by Devereux et al19: LV (grams) ⫽ 0.80 ⫻ 1.04 [(VSTd ⫹ LVIDd ⫹ PWTd)3 ⫺ (LVIDd)3] ⫹ 0.6, where thickness and dimension measurements are expressed in centimeters. LV relative wall thickness was calculated as twice the ratio of PWTd to LVIDd.4,20 Echocardiograms of adequate quality for performing M-mode measurements of the left ventricle were available in 1,488 men and 2,322 women. As previously described, ⱖ1 LV measurements could not be made in 34% of the elderly CHS cohort, with age being the strongest correlate of missing data.21 After excluding participants with prevalent disease at baseline, 884 men and 1,622 women remained and formed the basis for this study. Explanatory echocardiographic variables of interest included LV mass, LV internal dimensions in diastole and systole, ventricular septal thickness in diastole, LV posterior wall thickness in diastole, left atrial dimension, and mitral annular calcium (yes/no). In addition, a categorical variable for LV geometry and/or hypertrophy was defined, based on abnormal categories of LV relative wall thickness,4,20 and LV mass index (ratio of observed LV mass to predicted LV mass based on height and gender). Predicted LV mass was based on the results of a regression of observed LV mass (log transformed to increase normality) on height and gender.22 Abnormal LV relative wall thickness was defined to be any value more than the 95th percentile (⬎0.48); abnormal LV mass index was defined to be any value more than the 95th percentile (⬎1.635 predicted). LV geometry categories are defined in Table 1. Statistical analysis: Because baseline echocardiographic measurements were missing for many participants without prevalent disease who would otherwise have been included in the model (n ⫽ 1,274 missing values for most echocardiographic variables, but only 216 for left atrial dimension), incidence rates (per 1,000 person–years at risk) were compared between those with and without echocardiographic measurements. For all continuous predictor variables, age and cohort-adjusted means stratified by gender and incident disease status were calculated for each outcome of interest. Analysis of covariance was used to investigate which differences were significant. Cox regression was used to model the time to incident disease as a function of each echocardiographic variable individually. Time-to-event plots (stratified by the levels of each predictor of interest) were examined, and there was no evidence to indicate MAY 1, 2001
TABLE 2 Age and Cohort Adjusted Means of M-Mode Echocardiographic Variables by Incident Disease Status (excluding participants with prevalent disease at baseline) Incident CHD Echocardiographic Measure LV mass (g) (n ⫽ 2,506) Women Men LV dimension (diastole, mm) (n ⫽ 2,506) Women Men LV dimension (systole, mm) (n ⫽ 2,492) Women Men Ventricular septal thickness (diastole, mm) (n ⫽ 2,506) Women Men LV posterior wall thickness (diastole, mm) (n ⫽ 2,506) Women Men Left atrial dimension (mm) (n ⫽ 3,564) Women Men
Incident CHF
Incident Stroke
Mortality (All-cause)
⫹
o
⫹
o
⫹
o
⫹
o
136.5* 180.4†
130.0 167.9
152.4‡ 198.7‡
129.3 167.6
140.8* 182.8
130.5 169.6
135.7 183.1†
130.6 168.1
47.4 51.7
47.0 51.6
49.0 53.5†
46.8 51.3
48.4* 50.9
46.8 51.5
46.7 51.9
47.0 51.3
27.4 31.4
26.7 31.6
29.5‡ 33.5‡
26.5 30.4
28.0* 30.8
26.6 30.7
26.7 31.9*
26.6 30.4
8.7 9.7†
8.4 9.2
9.2‡ 10.1‡
8.5 9.2
8.6 9.9*
8.5 9.2
8.7 9.7†
8.5 9.2
8.4* 9.1†
8.1 8.8
8.6† 9.4†
8.1 8.8
8.3 9.4†
8.2 8.8
8.3 9.2†
8.2 8.8
37.0 39.4
38.8† 41.4†
37.0 39.3
37.7 39.9
‡
38.2 39.9
37.1 39.5
37.1 39.7
37.2 39.5
*p ⬍0.05; †p ⬍0.01; ‡p ⬍0.001 based on analysis of variance for equality of echocardiographic measurement means between disease status groups with gender.
that the proportional hazards assumption was inappropriate. Both unadjusted models and models including other cardiovascular risk factors are presented. Risk factors included age, gender, high-density lipoprotein cholesterol, weight, height, smoking status (current, former, never), hypertension status, and diabetes (normal, glucose intolerant, or diabetic) based on the new American Diabetes Association criteria. The echocardiographic variables were entered first as continuous variables, and then divided into gender-specific quartiles to capture possible nonlinearities in the relations. All models were checked for interactions between gender and echocardiographic measurements of interest. SPSS software (version 7.5, SPSS Inc., Chicago, Illinois) was used for all statistical analyses.
RESULTS
Missing echocardiographic measurements: CHS participants with missing echocardiographic LV measurements demonstrated a significantly higher rate of all-cause mortality (33.9 vs 16.5 per 1,000 person– years at risk) than those in whom echocardiographic measurements could be performed. Of importance, however, there were no differences in the rate of incident CHD, CHF, or stroke between the 2 groups.
Echocardiographic measurements per incident disease and gender (Table 2): When age-adjusted means
of M-mode echocardiographic variables were considered by gender, all measurements, in all incident disease status subgroups, were larger in men than women. As previously shown for LV mass, this difference in CHS was only partially explained by the larger body size of the men.21 When incident disease
status subgroups were considered, LV mass was significantly greater in men and women with, versus without, incident CHD and incident CHF; in women with, versus women without, stroke; and in men with, versus men without, death from all causes. There was a 23- to 31-g mean difference in age-adjusted mass in women and men with, versus without incident CHF. LV diastolic and systolic dimensions were greater in men and women with, versus without incident CHF (by 2.2 to 3.1 mm, on average), and greater in women with, versus without, incident stroke. LV systolic dimension was greater in men who died from all causes. Ventricular septal thickness was greater in men and women with, versus without, incident CHF and in men with, versus without, incident CHD, stroke, and allcause mortality. LV posterior wall thickness was greater in men and women with, versus without, incident CHD and CHF, and greater in men with, versus without, incident stroke and all-cause mortality. Finally, left atrial dimension was larger in men and women with, versus without, incident CHF. Echocardiographic measurement quartiles and allcause mortality: Unadjusted LV mass, ventricular sep-
tal thickness, posterior wall thickness, and left atrial dimension all showed a significant positive relation with all-cause mortality (range of hazards ratios for the highest vs lowest quartiles 1.15 to 1.80). Presence of mitral annular calcium also conferred increased risk for all-cause mortality (hazards ratio 1.71, p ⬍0.001). However, after adjustment for traditional risk factors, there remained significant relations only between allcause mortality and 2 echocardiographic variables:
CORONARY ARTERY DISEASE/ECHOCARDIOGRAPHIC PREDICTORS OF CARDIOVASCULAR EVENTS
1053
TABLE 3 Associations of Echocardiographic Measures and Incident Coronary Heart Disease (CHD) Incident CHD Echocardiographic Measure LV mass (g) Quartiles 1 2 3 4 LV dimension (diastole, mm) Quartiles 1 2 3 4 LV dimension (systole, mm) Quartiles 1 2 3 4 Ventricular septal thickness (diastole, mm) Quartiles 1 2 3 4 LV posterior wall thickness (diastole, mm) Quartiles 1 2 3 4 Left atrial dimension (mm) Quartiles 1 2 3 4 Mitral annular calcium No (n ⫽ 3,429) Yes (n ⫽ 244) LV anatomy category Height/gender adjustment Normal (n ⫽ 2,318) Concentric remodeling (n ⫽ 110) Eccentric LV hypertrophy (n ⫽ 55) Concentric LV hypertrophy (n ⫽ 14)
No. of Incident CHD Cases
Rate per 1,000 Person–Years
Unadjusted Hazards Ratio (95% CI)
Hazards Ratio Adjusted for Risk Factors (95% CI)
57 75 71 89
16.6 21.9 20.6 26.6
p ⬍0.001 1.00 1.34 (0.95–1.89) 1.25 (0.89–1.78) 1.63 (1.17–2.27)
p ⫽ 0.0053 1.00 1.33 (0.94–1.89) 1.22 (0.85–1.74) 1.43 (0.99–2.05)
66 72 76 78
19.9 21.6 22.0 21.7
p ⫽ 0.010 1.00 1.08 (0.78–1.51) 1.10 (0.79–1.53) 1.08 (0.78–1.50)
p ⫽ 0.9672 1.00 1.11 (0.79–1.55) 1.12 (0.80–1.57) 1.04 (0.74–1.47)
71 69 63 89
21.2 20.1 18.3 26.2
p ⫽ 0.001 1.00 0.95 (0.68–1.32) 0.86 (0.62–1.21) 1.23 (0.90–1.69)
p ⫽ 0.0775 1.00 0.89 (0.64–1.25) 0.89 (0.63–1.25) 1.24 (0.90–1.71)
65 68 76 83
18.2 19.7 22.6 25.2
p ⬍0.001 1.00 1.08 (0.77–1.52) 1.25 (0.90–1.74) 1.40 (1.01–1.93)
p ⫽ 0.0043 1.00 1.08 (0.76–1.52) 1.19 (0.85–1.67) 1.19 (0.85–1.67)
64 59 86 83
17.4 17.3 24.8 26.4
p ⬍0.001 1.00 1.00 (0.70–1.42) 1.44 (0.04–1.99) 1.54 (1.11–2.13)
p ⫽ 0.0021 1.00 0.94 (0.66–1.34) 1.32 (0.95–1.84) 1.33 (0.94–1.86)
93 115 115 113
18.3 23.6 23.8 24.0
408 47
21.7 35.4
p ⬍0.001 1.00 1.29 (0.98–1.70) 1.31 (0.99–1.72) 1.32 (1.00–1.74) p ⫽ 0.0015 1.00 1.63 (1.21–2.21)
p ⫽ 0.2402 1.00 1.24 (0.94–1.63) 1.21 (0.92–1.61) 1.21 (0.91–1.61) p ⫽ 0.0290 1.00 1.41 (1.04–1.93)
258 18 13 2
33.7 47.9 36.4 26.4
p ⫽ 0.0029 1.00 1.69 (1.05–2.73) 2.39 (1.37–4.18) 1.88 (0.47–7.54)
p ⫽ 0.0246 1.00 1.57 (0.97–2.54) 2.05 (1.16–3.62) 1.61 (0.40–6.54)
The p value for each echocardiographic variable represents the overall significance of the continuous variable in the model. CI ⫽ confidence interval.
LV mass (hazards ratio 1.21, p ⬍0.0122) and posterior wall thickness (hazards ratio 1.33, p ⬍0.0175). Of interest, there were positive relations of eccentric LV hypertrophy and concentric remodeling with all-cause mortality. There were no gender differences in associations reported in the following between echocardiographic variables and incident all-cause mortality, CHD, stroke, or CHF. Echocardiographic measurements and incident coronary heart disease (Table 3): Before adjustment for
traditional risk factors, all echocardiographic variables conferred an increased risk of incident CHD. Even after adjustment, the highest (versus lowest) quartiles of LV mass, septal thickness, posterior wall thickness, and the presence of mitral annular calcium conferred 1054 THE AMERICAN JOURNAL OF CARDIOLOGY姞
VOL. 87
increased risks for incident CHD (range of hazards ratios 1.2 to 1.4). Furthermore, the eccentric and concentric LV hypertrophy and concentric remodeling patterns conferred increased hazards ratios (range 1.6 to 2.0) for incident CHD, after adjustment for risk factors. Figure 1 illustrates the percent of subjects who were free of incident CHD, stratified by genderspecific quartiles of LV mass. Those in the lowest LV mass quartile had the longest CHD-free survival, with approximately 90% of participants free of incident CHD by the end of the study period— compared with 80% free of CHD in the highest quartile. The middle 2 quartiles were intermediate in terms of CHD-free survival. None of the curves were signifiMAY 1, 2001
DISCUSSION
FIGURE 1. Kaplan-Meier curves for incident CHD by LV mass gender-specific quartiles. The graph displays cumulative incident CHD-free survival (in days) in the 4 gender-specific LV mass quartiles. The vertical axis indicates the percent of the original study participants (who did not have a history of CHD) who remained free of incident CHD by each time point. The numbers 1 to 4 indicate the lowest to highest gender-specific quartiles of LV mass.
cantly different from one another (although there was a linear trend indicated by the overall p value of 0.0053). Echocardiographic measurements and incident stroke: Before adjustment for traditional risk factors,
LV mass, septal and posterior wall thickness, left atrial dimension, and the presence of mitral annular calcium conferred a significantly increased risk of incident stroke. However, after adjustment for risk factors, only echocardiographic LV mass conferred an increased risk (hazards ratio 1.52, highest vs lowest quartile) for incident stroke. None of the LV geometry categories conferred an increased risk for incident stroke, possibly because of the small number of cases in the 2 LV hypertrophy groups. Echocardiographic measurements and incident congestive heart failure (Table 4): All echocardiographic
variables (unadjusted for risk factors) were positively associated with incident CHF (range of unadjusted hazards ratios 1.9 to 4.1). Significant hazards ratios (e.g., 3.4 for LV mass) persisted for all of these echocardiographic variables, except for mitral annular calcium, after adjustment for risk factors. Both concentric and eccentric hypertrophy were associated with increased hazards ratios for incident CHF (3.3 and 3.0, respectively, after risk factor adjustment). Figure 2 illustrates the percent of the cohort who were free of incident CHF stratified by gender-specific quartiles of LV mass. The lowest 2 quartiles had equivalent curves, with approximately 97% free of incident CHF by the end of the study. In contrast, both the third, and particularly the highest quartile of mass, exhibited significantly worse curves than the 2 lowest quartiles, with 92% and 86% free of incident CHF, respectively, by the end of the study.
Present study findings: The present study demonstrates, in the elderly event-free CHS cohort, the predictive value of 2-dimensional M-mode echocardiographic measurements. Specifically, relations remained significant, even after adjustment for traditional risk factors, between: (1) incident CHF and LV mass, internal diastolic and systolic dimensions, septal and posterior wall thickness, concentric and eccentric LV hypertrophy, and left atrial dimension; (2) incident CHD and LV mass, septal and posterior wall thickness, eccentric hypertrophy and mitral annular calcium, and (3) all-cause mortality and LV mass, posterior wall thickness, and non-normal ventricular geometry category. These data extend previous observations, because unlike the Framingham Heart Study, CHS is a multicenter study, specifically of the elderly, and includes a relatively large cohort of blacks, as well as whites. Ventricular mass and incident events: In CHS, LV mass was linearly associated with all-cause mortality even after adjustment for traditional risk factors. More striking, the hazards ratios in CHS for incident CHF for the third and fourth quartiles of adjusted mass were 2.0 and 3.4, respectively (compared with the lowest quartile). In contrast, the hazards ratios for incident CHD in CHS were less than relative risks (7.8 in men and 3.4 in women) for CHD events conferred by the highest versus the lowest quartile of risk factor-adjusted mass reported in a 4-year follow-up in the original cohort of the Framingham study (mean age 69 years).2 The explanation for the different magnitudes of hazards ratios between the 2 studies is unclear. Ventricular geometry and incident events: Increased LV internal dimensions in the CHS cohort were positively associated with incident CHF, similar to previous study from Framingham investigators.6,7 In addition, in CHS, ventricular septal and posterior wall thickness were positively associated with incident CHF. Eccentric LV hypertrophy and concentric remodeling, but not concentric hypertrophy, conferred increased hazards ratios for incident CHD, even after adjustment for traditional risk factors. In contrast, Koren et al4 reported that in patients with uncomplicated essential hypertension, concentric and eccentric hypertrophy were importantly related to the incidence of CVD events (p ⬍0.03) and mortality (p ⬍0.001). LV mass and geometry stratified risk independently of, and more strongly than, blood pressure or other potentially reversible risk factors. Similarly, among Framingham subjects ⱖ40 years old who were free of clinical CVD, those with concentric hypertrophy had the worst prognosis, followed by those with eccentric hypertrophy, concentric remodeling, and normal geometry.23 Concentric hypertrophy may not have reached statistical significance as a predictor of incident CHD in CHS because of the small number (n ⫽ 2) of incident CHD cases. Ventricular geometry, mass, and stroke: In CHS, after adjustment for traditional risk factors, LV mass was associated with a modestly increased hazards ratio (1.52 in the highest quartile) for stroke. No LV
CORONARY ARTERY DISEASE/ECHOCARDIOGRAPHIC PREDICTORS OF CARDIOVASCULAR EVENTS
1055
TABLE 4 Associations of Echocardiographic Measurements and Incident Congestive Heart Failure (CHF) Incident CHF Echocardiographic Measure LV mass (g) Quartiles 1 2 3 4 LV dimension (diastole, mm) Quartiles 1 2 3 4 LV dimension (systole, mm) Quartiles 1 2 3 4 Ventricular septal thickness (diastole, mm) Quartiles 1 2 3 4 LV posterior wall thickness (diastole, mm) Quartiles 1 2 3 4 Left atrial dimension (mm) Quartiles 1 2 3 4 Mitral annular calcium No (n ⫽ 3,428) Yes (n ⫽ 244) LV anatomy category Height/gender adjustment Normal (n ⫽ 2,320) Concentric remodeling (n ⫽ 11) Eccentric LV hypertrophy (n ⫽ 52) Concentric LV hypertrophy (n ⫽ 13)
No. of Incident CHF Cases
Rate per 1000 Person–Years
Unadjusted Hazards Ratio (95% CI)
Hazards Ratio Adjusted for Risk Factors (95% CI)
19 19 41 73
5.3 5.3 11.5 21.6
p ⬍0.001 1.00 1.00 (0.53–1.88) 2.18 (1.27–3.76) 4.11 (2.48–6.81)
p ⬍0.001 1.00 1.00 (0.53–1.89) 2.01 (1.16–3.51) 3.36 (1.96–5.74)
26 33 36 57
7.6 9.5 10.1 15.7
p ⬍0.001 1.00 1.25 (0.75–2.09) 1.32 (0.80–2.19) 2.05 (1.29–3.27)
p ⬍0.001 1.00 1.33 (0.79–2.23) 1.44 (0.86–2.41) 2.05 (1.26–3.33)
28 24 30 70
7.9 6.7 8.5 20.4
p ⬍0.001 1.00 0.84 (0.49–1.45) 1.07 (0.64–1.80) 2.57 (1.66–3.99)
p ⬍0.001 1.00 0.80 (0.46–1.39) 1.20 (0.71–2.02) 2.69 (1.71–4.23)
22 29 46 55
6.0 8.1 13.3 16.2
p ⬍0.001 1.00 1.37 (0.79–2.38) 2.25 (1.36–3.74) 2.76 (1.69–4.53)
p ⬍0.001 1.00 1.22 (0.70–2.12) 1.87 (1.12–3.13) 1.87 (1.13–3.11)
24 28 42 58
6.3 8.0 11.7 18.0
p ⬍0.001 1.00 1.28 (0.74–2.20) 1.88 (1.14–3.11) 2.94 (1.83–4.73)
p ⬍0.001 1.00 1.10 (0.63–1.90) 1.47 (0.88–2.45) 2.00 (1.22–3.27)
40 52 60 80
7.6 10.3 12.0 16.7
216 29
11.2 21.2
p ⬍0.001 1.00 1.35 (0.90–2.04) 1.57 (1.06–2.35) 2.21 (1.51–3.23) p ⫽ 0.0012 1.00 1.89 (1.29–2.79)
p ⬍0.001 1.00 1.28 (0.84–1.94) 1.49 (0.99–2.24) 1.90 (1.28–2.82) p ⫽ 0.1454 1.00 1.34 (0.90–2.00)
129 9 11 3
9.8 15.9 40.6 57.0
p ⬍0.001 1.00 1.66 (0.84–3.26) 4.27 (2.31–7.91) 6.26 (1.99–19.67)
p ⫽ 0.0018 1.00 1.40 (0.71–2.76) 2.95 (1.56–5.57) 3.32 (1.04–10.60)
The p value for each echocardiographic variable represents the overall significance of the continuous variable in the model. Abbreviation as in Table 3.
geometric pattern conferred an increased adjusted hazards ratios for incident stroke. However, geometric subgroup analyses were limited due to the small numbers of events–for example, there were only 2 incident stroke cases in the concentric hypertrophy subgroup. Recent studies have documented the association of carotid plaque and stenosis, and increased intimalmedial thickness, with higher LV mass.24,25 Left atrial dimension and incident events: In CHS, left atrial dimension was predictive of all-cause mortality, incident stroke, and CHD; however, these associations became nonsignificant after adjustment for traditional risk factors. In contrast, left atrial dimension, after adjustment, remained predictive of incident CHF. These results differ from both the Framingham 1056 THE AMERICAN JOURNAL OF CARDIOLOGY姞
VOL. 87
Heart Study8 and the Stroke Prevention in Atrial Fibrillation trial,9,10 in which left atrial dimension was related to strokes and death in the former, and to thromboembolic events (e.g., strokes and transient ischemic attacks) in the latter. Different findings in these 3 studies may relate to differences in subject ages and selection criteria. Mitral annular calcium and incident events: In CHS, mitral annular calcium conferred an increased unadjusted hazards ratio for all-cause mortality, incident CHD, stroke, and CHF, but remained predictive only of incident CHD after adjustment for risk factors. In contrast, in the Framingham study,11 mitral annular calcium conferred a greater risk for incident stroke than did atrial fibrillation. There has been speculation MAY 1, 2001
FIGURE 2. Kaplan-Meier curves for incident CHF by LV mass gender-specific quartiles. The graph displays cumulative incident CHF-free survival (in days) in the 4 gender-specific LV mass quartiles. The format is similar to that for Figure 1.
regarding the pathophysiologic explanation for the relation of mitral annular calcium and various incident CVD events. Similar risk factors may be involved in genesis of both mitral annular calcium and CHD. This hypothesis is strengthened by the association found in CHS between coronary risk factors and aortic valve sclerosis.26 In summary, this study suggests that in an elderly population, selected M-mode echocardiographic measurements are subclinical markers that confer independent prognostic information for incident CVD events, particularly for CHD and CHF.13 1. Casale PN, Devereux RB, Milner M, Zullo G, Pickering TG, Laragh JH. Value
of echocardiographic measurement of left ventricular geometry mass in predicting cardiovascular morbid events in hypertensive men. Ann Intern Med 1986; 105:173–178. 2. Levy D, Garrison RJ, Savage DD, Kannell WB, Castelli WP. Left ventricular geometry, mass and incidence of coronary heart disease in an elderly cohort. The Framingham Heart Study. Ann Intern Med 1989;110:101–107. 3. Levy D, Garrison RJ, Savage DD, Kannell WB, Castelli WP, Laragh JH. Prognostic implications of echocardiographically determined left ventricular geometry and mass in the Framingham Heart Study. N Engl J Med 1990;322:1561– 1566. 4. Koren MJ, Devereux RB, Casale PN, Savage DD. Relation of left ventricular geometry mass and geometry to morbidity and mortality in uncomplicated essential hypertension. Ann Intern Med 1991;114:345–352. 5. Ghali JK, Liao Y, Simmons B, Castaner A, Cao, G, Cooper RS. The prognostic role of left ventricular geometry hypertrophy in patients with or without coronary artery disease. Ann Intern Med 1992;117:831– 836. 6. Lauer MS, Evans JC, Levy D. Prognostic implications of subclinical left
ventricular dilatation and systolic dysfunction in men free from overt cardiovascular disease (the Framingham Heart Study). Am J Cardiol 1992;70:1180 –1184. 7. Vasan RS, Larson MG, Benjamin EJ, Evans JC, Levy, D. Left ventricular dilatation and the risk of congestive heart failure in people without myocardial infarction. N Engl J Med 1997;336:1350 –1355. 8. Benjamin EJ, D’Agostino RB, Belanger AJ, Wolf PA, Levy D. Left atrial size and the risk of stroke and death: the Framingham Heart Study. Circulation 1995;92:835– 841. 9. The Stroke Prevention in Atrial Fibrillation Investigators. Predictors of thromboembolism in atrial fibrillation: I. Clinical features of patients at risk. Ann Intern Med 1992;116:1–5. 10. The Stroke Prevention in Atrial Fibrillation Investigators. Predictors of thromboembolism in atrial fibrillation: II. echocardiographic features of patients at risk. Ann Intern Med 1992;116:6 –12. 11. Benjamin EJ, Plehn JF, D’Agostino RB, Belanger AJ, Comai K, Fuller, DL, Wolf, PA, Levy D. Mitral annular calcification and the risk of stroke in an elderly cohort. N Engl J Med 1992;327:374 –379. 12. Devereux RB, Alderman MH. Role of preclinical cardiovascular disease in the evolution from risk factors exposure to development of morbid events. Circulation 1993;88:1444 –1455. 13. Fried L, Borhani N, Enright P, Furberg CD, Gardin JM, Kronmal RA, Kuller LH, Manolio TA, Mittelmark MB, Newman A. Cardiovascular Health Study. Design and rationale. The CHS Collaborative Research Group. Ann Epidemiol 1991;1:263–276. 14. Tell G, Fried L, Hermanson B, Manolio TA, Newman AB, Borhani N, for the Cardiovascular Health Study Collaborative Research Group. Recruitment of adults 65 years and older as participants in the Cardiovascular Health Study. Ann Epidemiol 1993;3:358 –366. 15. Psaty B, Kuller LH, Bild D, Burke GL, Kittner SJ, Mittlemark M, Price TR, Rautaharju PM, Robbins J. Methods of assessing prevalent cardiovascular disease in the Cardiovascular Health Study. Ann Epidemiol 1995;5:270 –277. 16. Ives DG, Fitzpatrick AL, Bild DE, Psaty BM, Kuller KH, Crowley PM, Cruise RG, Theroux S. Surveillance and ascertainment of cardiovascular events: the Cardiovascular Health Study. Ann Epidemiol 1995;5:278 –285. 17. Gardin JM, Wong ND, Bommer W, Klopfenstein HS, Smith VE, Tabatznik B, Siscovick D, Lobodzinski S, Anton-Culver H, Manolio TA. Echocardiographic design of a multi-center investigation of free-living elderly subjects: the Cardiovascular Health Study. J Am Soc Echocardiogr 1992;5:63–72. 18. Sahn DJ, DeMaria A, Kisslo J, Weyman A, the Committee on M-mode Standardization of the American Society of Echocardiography. Recommendations regarding quantitation in M-mode echcoardiography: results of a survey of echocardiographic methods. Circulation 1978;58:1072–1083. 19. Devereux RB, Alonso DR, Lutas EM, Gottlieb GJ, Campo E, Sachs I, Reichek N. Echocardiographic assessment of left ventricular geometry hypertrophy: comparison with necropsy findings. Am J Cardiol 1986;57:450 – 458. 20. Ganau A, Devereux RB, Roman MJ, de Simone G, Pickering TG, Saba PS, Vargiu P, Simongini I, Laragh JH. Patterns of left ventricular geometry hypertrophy and geometric remodelling in essential hypertension. J Am Coll Cardiol 1992;19:1550 –1558. 21. Gardin JM, Siscovick D, Anton-Culver H, Lynch JC, Smith VE, Klopfenstein HS, Bommer WJ, Fried L, O’Leary D, Manolio TA. Sex, age and disease affect echocardiographic left ventricular geometry, mass, and systolic function in the free-living elderly: the Cardiovascular Health Study (CHS). Circulation 1995; 91:1739 –1748. 22. Lauer MS, Larson MG, Levy D. Gender-specific reference M-mode values in adults: population-derived values with consideration of the impact of height. J Am Coll Cardiol 1995;26:1039 –1046. 23. Krumholz HM, Larson M, Levy D. Prognosis of left ventricular geometric patterns in the Framingham Heart Study. J Am Coll Cardiol 1995;25:879 – 884. 24. Roman MJ, Pickering TG, Schwartz JE, Pini R, Devereux RB. Association of carotid atherosclerosis and left ventricular hypertrophy. J Am Coll Cardiol 1995;25:83–90. 25. Kronmal RA, Smith VE, O’Leary DH, Polak JF, Gardin, Manolio TA. Carotid artery measures are strongly associated with left ventricular mass in older adults: a report from the Cardiovascular Health Study. Am J Cardiol 1996;77: 628 – 633. 26. Stewart BF, Siscovick D, Lind BK, Gardin JM, Gottdiener JS, Smith VE, Kitzman DW, Otto CM. Clinical factors associated with calcific aortic valve disease. J Am Coll Cardiol 1997;29:630 – 634.
CORONARY ARTERY DISEASE/ECHOCARDIOGRAPHIC PREDICTORS OF CARDIOVASCULAR EVENTS
1057