Author + information
- Received February 26, 2015
- Revision received April 21, 2015
- Accepted April 23, 2015
- Published online September 1, 2015.
- Khiet Hoang, MD∗,
- Yanglu Zhao, MD, MS∗,
- Julius M. Gardin, MD, MBA†,
- Mercedes Carnethon, PhD‡,
- Ken Mukamal, MD§,
- David Yanez, PhD‖ and
- Nathan D. Wong, PhD∗∗ ()
- ∗Heart Disease Prevention Program, Division of Cardiology, Department of Medicine, University of California, Irvine, Irvine, California
- †Department of Medicine, Hackensack University Medical Center, Hackensack, New Jersey
- ‡Department of Preventive Medicine, Northwestern University, Chicago, Illinois
- §Department of Medicine, Harvard University, Boston, Massachusetts
- ‖Department of Biostatistics, University of Washington, Seattle, Washington
- ↵∗Reprint requests and correspondence:
Dr. Nathan D. Wong, Heart Disease Prevention Program, Division of Cardiology, Department of Medicine, C240 Medical Sciences, University of California, Irvine, California 92697.
Objectives The purpose of this study was to examine the prognostic significance of left ventricular (LV) mass for cardiovascular disease (CVD) events in older adults with and without metabolic syndrome (MetS) and diabetes mellitus (DM).
Background MetS and DM are associated with increased CVD risk, but it is unclear in these groups whether subclinical CVD as shown by increased LV mass improves risk prediction compared to standard risk factors in older individuals.
Methods We studied 3,724 adults (mean 72.4 ± 5.4 years of age, 61.0% female, 4.4% African-American) from the Cardiovascular Health Study who had MetS but not DM or had DM alone or had neither condition. Cox regression was used to examine the association of LV mass, (alone and indexed by height and body surface area [BSA]) as determined by echocardiography, with CVD events, including coronary heart disease (CHD), stroke, heart failure (HF), and CVD death, as well as total mortality. We also assessed the added prediction, discriminative value, and net reclassification improvement (NRI) for clinical utility of LV mass compared to standard risk factors.
Results Over a mean follow-up of 14.2 ± 6.3 years, 2,180 subjects experienced CVD events, including 986 CVD deaths. After adjustment for age, sex and standard risk factors, LV mass was positively associated with CVD events in those with MetS (hazard ratio [HR]: 1.4, p < 0.001) and without MetS (HR: 1.4, p < 0.001), but not DM (HR: 1.0, p = 0.62), with similar findings for LV mass indexed for height or BSA. Adding LV mass to standard risk factors moderately improved the prediction accuracy in the overall sample and MetS group from changes in C-statistics (p < 0.05). Categorical-free net reclassification improvement increased significantly by 17% to 19% in those with MetS. Findings were comparable for CHD, CVD mortality, and total mortality.
Conclusions LV mass is associated with increased CVD risk and provides modest added prediction and clinical utility compared to standard risk factors in older persons with and without MetS but not with DM.
Older persons with metabolic syndrome (MetS) and diabetes mellitus (DM) are more likely to have subclinical atherosclerosis and are at greater risk of cardiovascular disease (CVD) events (1–3) and mortality (4). Previous studies have identified left ventricular (LV) mass independently predicts CVD events (5–7). Although the association of MetS (5) and the number of MetS risk factors (7) with LV mass has been demonstrated, and DM adversely impacts hypertrophic remodeling through increased LV mass and larger cavity dimensions (8), there are limited data examining the value of LV mass for predicting CVD events in persons with MetS or DM. Although a smaller previous study compared the prognosis of increased LV mass in diabetic and nondiabetic hypertensive individuals (9), to our knowledge, no population-based study has compared the prognostic significance of LV mass in persons with and without MetS and DM. Although DM is a well-known risk factor for coronary heart disease (CHD), it has been shown by some studies to confer a lower risk of subsequent cardiac complications than CHD (10). There is a need to better identify what other screening methods for subclinical CVD can further improve risk prediction in persons with MetS and DM (11). For instance, it is known such persons demonstrate a greater extent of myocardial ischemia (12) and coronary calcium (13,14), with the latter providing prognostic value for CVD events (15). Whether subclinical CVD as shown by higher LV mass provides significant incremental prognostic value in predicting CVD events compared to standard risk factors under these conditions is unclear, especially in those with MetS and DM and in older persons who have a longer exposure to these conditions. Such information would be useful to judge the utility of LV mass assessment in these groups.
This study examined whether readily available echocardiographic measurements of LV mass added to standard CVD risk factors in the prediction of CVD events in older persons with and without MetS or DM. Our analysis addressed the question of whether there is a role for these readily available measurements in risk stratification for these populations.
Study sample and recruitment
Our analyses included 3,724 adults 65 to 95 years of age from the Cardiovascular Health Study (CHS), a prospective U.S. National Institutes of Health-sponsored study of older adults, which studied risk factors and subclinical measurements of CVD and their outcomes. Initial enrollment from 1989 to 1990 recruited 5,201 participants, whereas a second cohort of 687 African-American participants was recruited from 1992 to 1993. Specifically, of the initial cohort of 5,201 subjects, the current analysis included CHS participants who had baseline measurements of LV mass from 2-dimensionally directed M-mode echocardiography as well as information for incidence of CVD events; patients with prior CVD events were excluded. Participants were initially recruited from Heath Care Financing Administration Medicare eligibility lists and other household members from 4 U.S. geographic regions: Forsyth County, North Carolina; Sacramento County, California; Washington County, Maryland; and Pittsburgh, Pennsylvania. Participant consent was obtained during baseline examination. Baseline examination data consisted of medical history, physical examination, and fasting blood analyses. The methodology and design of CHS have been previously reported (15). Up to 22 years of follow-up data were available through June 30, 2004, with vital status known for all 3,724 subjects included in the study, with complete risk factor data (no persons lost to follow-up). This project was exempt from Institutional Review Board review due to the use of de-identified data.
Risk factors were measured by standardized methodology, as previously described, and included systolic and diastolic blood pressure (BP), low-density lipoprotein cholesterol (LDL-C) and high-density lipoprotein-cholesterol (HDL-C), triglycerides, glucose, waist circumference, and body mass index (BMI) (15). Subjects were classified as having MetS without DM (according to American Heart Association/National Heart, Lung, and Blood Institute [AHA/NHLBI] criteria], or DM, or neither condition. MetS (n = 1,178) without DM was defined, according to the AHA/NHLBI definition (16), as having any 3 of the following 5 criteria: elevated BP (≥130 systolic or ≥85 mm Hg) or treatment for hypertension; low HDL-C (<40 mg/dl in males or <50 mg/dl in females); elevated triglycerides (≥150 mg/dl); increased waist circumference (>88 cm [35 inches] in females or >102 cm [40 inches] in males); or impaired fasting glucose (100 to 125 mg/dl). DM (n = 485 subjects) was defined as having a fasting glucose concentration of ≥6.9 mmol/l (126 mg/dl), taking oral hypoglycemic medication, or self-reported use of insulin. Subjects with neither condition (n = 2,061) were also included in our analyses.
The protocol for performing and reading transthoracic echocardiograms has been previously described (17). Briefly, a baseline echocardiogram was recorded onto super-VHS tape by using a standardized protocol, with measurements made at the Echocardiography Reading Center at University of California, Irvine, from 1989 to 1990, and at Georgetown University from 1992 to 1993 from digitized images, using an off-line image analysis system equipped with customized computer algorithms. Quality control measurements included standardized training of sonographers and readers, periodic observation of a sonographer by a trained echocardiographer, and blind duplicate readings to establish inter-reader and intrareader measurement variability. This study focused on 2-dimensionally directed M-mode measurements of LV mass, which were calculated as described by Devereux et al. (18), where LV mass (g) = 0.80 × 1.04 [(VSTd + LVIDd + PWTd)3 − (LVIDd)3] + 0.6 cm, where VSTd = ventricular septal thickness in diastole; LVIDd = LV internal dimension in diastole; and PWTd = posterior wall thickness in diastole. We also present our data according to LV mass indexed to height (cm1.7), given that this variable has been recently proposed to represent more accurate scaling (19) than older scaling (such as by a power of 2.7) as well as by body surface area (BSA).
CVD and CHD events were adjudicated by the CHS endpoints committee of physician investigators. Incident CVD was defined as CHD, stroke, heart failure (HF), or claudication, with CVD deaths due to either of these incident conditions. CHD events included incident nonfatal myocardial infarction, angina requiring hospitalization, coronary artery angioplasty, coronary bypass surgery, or death caused by “atherosclerotic CHD.” CHS criteria for angina required a report of symptoms such as chest pain, chest tightness, or shortness of breath; the diagnosis of angina from a physician; and being under medical treatment for angina (including nitroglycerin, beta blocker, or calcium channel blocker). Total CVD and CHD events, CVD mortality, and total mortality were defined as occurring after the baseline echocardiographic assessment of LV mass. The first occurrence of a qualifying event was used as the individual’s “event,” so recurrent events were not included in the analysis. Follow-up time was defined from the baseline LV mass echocardiographic assessment to the date of first occurrence of a CVD event (or CHD, CVD death, or total mortality for analyses specific to those endpoints).
Descriptive statistics of proportions for categorical variables and mean ± SD for continuous variables are presented by disease group and compared by using chi-square test of proportions or analysis of variance among groups, respectively. Cox proportional hazards regression analysis was used to examine the association of LV mass with time to the primary outcome of a first CVD event and with time to the secondary endpoints of CHD, CVD mortality, and total mortality, providing hazard ratios (HRs) and 95% confidence intervals (CIs). These analyses were adjusted for age, sex, ethnicity, and standard risk factors (systolic BP, diastolic BP, hypertensive medications, HDL-C, LDL-C, total cholesterol, triglycerides, lipid medications, BMI, and fasting glucose). LV mass was stratified according to sex-specific quartiles in grams and was also examined continuously according to SD of LV mass indexed by height (g/m1.7), SD of LV mass indexed by BSA (g/m2), and SD of LV mass of 30 g. The area under the receiving operator characteristic curve (AUC) was used to examine the incremental value of LV mass above standard risk factors for the prediction of CVD events. We constructed logistic regression models with and without LV mass measurements to compare AUC differences. In addition, to examine the added clinical utility of echocardiographic LV mass compared to standard risk factors, the category-free NRI was calculated as: [(number of events reclassified with higher risk − number of events reclassified with lower risk)/number of events] + [(number of nonevents reclassified as lower risk − number of nonevents reclassified as higher risk)/number of nonevents]. SAS statistical software version 9.4 (SAS Institute, Cary, North Carolina) (20) was used for analysis. A p value of <0.05 (and a p value of <0.1 for interaction test) was considered statistically significant.
The age of our 3,274 participants was 72.4 ± 5.4 years, with 61% females and 4.4% African Americans (Table 1). As expected, participants with MetS or DM had significantly higher mean BPs, lipid measurements, serum glucose, BMI, and LV mass than persons with neither disease. LV mass values for those with MetS, DM, or neither were 155.7 ± 28.3 g, 163.5 ± 31.6 g, and 142.4 ± 28.4 g, respectively (p < 0.0001). Mean LV mass values indexed to height (LV mass [g/m1.7]) for those with MetS, DM, or neither were 66.9 ± 9.8 g/m, 69.1 ± 11.4 g/m, and 60.8 ± 10.6 g/m, respectively (p < 0.0001). Mean LV mass values indexed to BSA (LV mass in g/BSA [m2]) for those with MetS, DM, or neither were 85.6 ± 10.2 g/m2, 89.0 ± 12.4 g/m2, and 82.8 ± 11.1 g/m2, respectively (p < 0.0001).
Cardiovascular disease events
Over a mean follow-up of 14.2 ± 6.3 years, 2,180 participants experienced at least 1 CVD event. Unadjusted rates of total CVD events per 1,000 person years were highest in those with DM. A stepwise increase of unadjusted rates of total CVD events per 1,000 person years was observed across quartiles of LV mass for those with MetS and DM and those with neither condition (Figure 1).
Relationship between LV mass and outcomes
Findings from adjusted Cox proportional hazards regression are shown in Table 2 for primary (total CVD) and secondary outcomes (total CHD, CVD mortality, and all-cause mortality). In participants with neither MetS nor DM and persons with MetS alone, higher LV mass and indexed LV mass were risk factors for total CVD, total CHD, CVD mortality, and all-cause mortality. These associations were not observed in persons with DM. Interaction test results for disease groups and LV mass were significant for total CHD (p = 0.049), CVD death (p = 0.057), and total mortality (p = 0.060); for disease groups and LV mass indexes, results were only significant for CVD death in relation to LV mass/height1.7 (p = 0.064) and for LV mass/BSA (p = 0.029). Similar findings were noted when subjects were stratified by sex. Both men and women with neither disease or with MetS alone had a higher risk of total CVD events per SD of increase in LV mass; men and women with DM did not (Table 3).
Although unadjusted HR values for CVD events in relation to LV mass (per SD) were significant (p < 0.01) in all 3 disease groups, they were weaker in those with DM (HR: 1.16) than in those with MetS (HR: 1.27) or neither condition (HR: 1.25), and in those with DM were further attenuated to being nonsignificant after adjustment for sex (higher LV mass and event rates in men with DM), age, systolic BP, and cholesterol in particular. In addition, the relationship of LV mass with CVD events did not differ between men and women (interaction test results were not significantly different). Because <5% of our subjects (n = 165) were African American, the sample size was insufficient to show relationships with CVD events in those with MetS or DM; however, in those with neither condition, risks of CVD events (per SD of LV mass) appeared to be greater in African Americans (HR: 2.42 [95% CI: 1.26 to 4.64], p < 0.01) than in whites (HR: 1.36 [95% CI: 1.22 to 1.51], p < 0.001) with similar findings for indexed LV mass measurements (results not shown).
Adjusted Cox proportional hazard models for outcomes were also examined by quartiles of LV mass. Participants with neither MetS nor DM and those with MetS alone who were in the highest quartile of LV mass had significantly increased risks for total CVD (HR: 1.9 [p < 0.0001] and 2.0 [p < 0.0001], respectively) compared to those in the first quartile (Figure 2). Similarly, the highest quartile (versus lowest quartile) of LV mass independently predicted secondary endpoints of total CHD (HR: 2.0, p < 0.0001), CVD death (HR: 2.4, p < 0.0001), and all-cause mortality (HR: 1.5, p < 0.01) in those with neither condition. Hazard ratios were increased for total CHD (HR: 1.6, p = 0.03) and CVD death (HR: 1.7, p = 0.04) but not for all-cause mortality (HR: 1.3, p = 0.13) in those with MetS in the highest versus lowest quartile of LV mass. In contrast, there was no significant increased risk of both primary and secondary endpoints across quartiles of LV mass in those with DM.
The AUC did not show significant incremental predictive value for total incident CVD events between the base model and the other 3 models for the prediction of CVD events with LV mass, LV mass/height1.7 or LV mass/BSA across all 3 disease groups, except modestly (p < 0.05) for LV mass/height1.7 and LV mass/BSA in those with MetS (Table 4). There was also a significant (p < 0.05) improvement in c-statistic in the overall sample comparing models, with LV mass added to those with risk factors alone, although the absolute degree of improvement was minimal (both were 0.63 to the second decimal). We additionally examined the AUC improvement for CVD mortality and total mortality. Results showed that among subjects with neither disease, AUC increased from 0.64 to 0.66 (p < 0.05) for all 3 LV mass scores. AUCs for total mortality ranged from 0.70 to 0.74 after including LV mass measurements in the model, but the improvement was not significant (data not shown).
Analysis results from NRI showed modest added clinical utility for prediction of CVD events and ranged from 4% to 7% in the nondisease group and 17% to 19% in the MetS group but was significant only in the MetS group (p < 0.01), comparing the base model and models with 3 forms of LV mass measurements; however, in those with DM, NRI was not significant (Table 5). In the overall sample, there was a significant NRI from 9% to 10% (p < 0.01). NRI values for CVD mortality were 9% to 15% in the 3 disease groups (p < 0.05 in MetS and no disease group for LV mass/height1.7 and LV mass/BSA). NRI values for total mortality were greatest in those patients with MetS (10%; p value was not significant), and was <5% in the other 2 groups (p = nonsignificant; data not shown).
Our study found that increased LV mass (highest quartile) was associated with increases in risk for total CVD events, total CHD events, and CVD and total mortality in those with and without MetS but not in those with DM. Our paper is the first to report added discriminative and clinical utility for echocardiographic LV mass over standard CVD risk factors using ROC and NRI techniques.
Our report corroborates earlier findings from shorter-term follow-up regarding the overall relationship of echocardiographic predictors (including LV mass) to CVD events in the entire CHS cohort by Gardin et al. (5). In addition, Kuller et al. (21) previously reported among persons with DM that the general presence of subclinical CVD (from the presence of a low ankle-brachial index, increased carotid intimal medial thickness or stenosis, major ECG abnormalities, or angina) was associated with a 2-fold greater risk of incident CHD. More recently, in the longitudinal MESA (Multiethnic Study of Atherosclerosis), LV mass measured by cardiac magnetic resonance imaging was shown to improve the c-statistic over traditional risk factors for the prediction of incident HF (22), although this relationship was not examined in those with MetS and DM.
DM is noted to have an adverse effect on hypertrophic remodeling through promotion of increases in LV mass and dimensions (7). The Framingham Heart study identified an association among DM and increased LV wall thickness and mass that was independent of traditional risk factors in women but not in men (23). LV hypertrophy is common in those with DM, but previous screening modalities such as ECG and N-terminal pro–B-type natriuretic peptide have been noted to be inadequate for detecting LV hypertrophy (24). Our data also suggest echocardiographic LV mass has limited utility to stratify risk in persons with DM, at least in older adults, which comprised our cohort. Although increased LV mass is a well-known marker of end-organ hypertensive damage, a possible explanation for our lack of a relationship in those with DM may be their high baseline risk, supported by the common notion that DM is a coronary risk equivalent (25) (which would be the case especially in our older cohort), and hypertension or other highly prevalent risk factors in older persons with DM may have obscured our relationships with LV mass in such persons. Of note, we observe that even though the unadjusted relation of LV mass with CVD events is significant in those with DM, it is of lower magnitude than those with MetS and those with neither condition and is attenuated to being nonsignificant after adjustment for age, sex (in particular), systolic BP, and cholesterol, whereas the LV mass relationship with CVD events remains significant after adjustment for these and other risk factors in those without DM. Alternatively, if smaller LV mass is protective, the prognostic value of LV mass in diabetics may be lost because of their higher baseline LV mass, particularly in a cohort of older subjects who have had years of exposure to DM and more advanced subclinical CVD. Other measurements that more directly reflect atherosclerosis burden may be more important for further risk stratification of the patient with DM, such as coronary calcium, which has been shown to add prognostic value in such patients (13,14). MetS, however, is a more heterogeneous condition associated with a wide variation in CVD risk (26), with many persons at intermediate risk, where further evaluation such as by echocardiographic LV mass may be helpful for risk stratification; our data support this by showing a modest added value for echocardiographic LV mass in risk prediction in such persons and in those without MetS.
Limitations of our study include the fact that the unidimensional nature of our M-mode measurements does not take into account changes in eccentricity based on long-axis and short-axis LV measurements; thus, future studies involving 2- or 3-dimensional echocardiographic recordings should investigate whether LV mass is erroneously estimated by M-mode echocardiography in conditions such as obesity, MetS, and DM, in which the ventricle may be more spherically shaped. In fact, Bluemke et al. (27) showed that stroke and CHD events were better predicted by abnormal LV geometry (e.g., increased LV mass-to-volume ratio), whereas HF events were driven primarily by increased LV mass alone. Our study did not have measurements of LV geometry or systolic or diastolic function, which are known to be affected in DM. However, other studies have shown that LV mass was the single 2-dimensional echocardiographic measurement consistently associated with total and individual CVD endpoints (28). Substantially higher overall CVD event rates in the diabetic group, which were not further increased by higher LV mass levels, contrast with those without DM, where increased LV mass added more to CVD event prediction. Additionally, most of our cohort was of Caucasian descent, consequently, our findings may not be generalizable to other racial/ethnic groups and to younger populations.
Our study shows that in older persons with MetS and in those without but not in those with DM, echocardiographic LV mass is positively associated with increases in total CVD risk, including CHD, HF, and stroke, and adds modest clinical utility for CVD prediction compared with standard risk factors. Thus, measurement of LV mass, although possibly useful to stratify risk in older persons without DM, may be of limited clinical utility in those with DM, who are already at significant CVD risk.
COMPETENCY IN MEDICAL KNOWLEDGE: Evaluation of left ventricular mass using 2-dimensionally guided M-mode echocardiography is known to provide risk stratification for future cardiovascular events beyond information provided by standard risk factors. This study confirms these findings in older adults generally and in those without diabetes, including those with and without metabolic syndrome. The weaker role of left ventricular mass for improving risk prediction in those with diabetes may be due to the important effect of other risk factors.
TRANSLATIONAL OUTLOOK: Older persons and especially those with diabetes have significant but often varied risks for developing cardiovascular events. Future studies might examine the roles of other structural and functional characteristics, especially measured by newer technologies such as cardiac magnetic resonance for further refining cardiovascular risk prediction in such patients.
This research was supported by National Institutes of Health/National Heart, Lung, and Blood Institute contracts HHSN268201200036C, HHSN268200800007C, N01 HC55222, N01HC85079, N01HC85080, N01HC85081, N01HC85082, N01HC85083, and N01HC85086 and grant HL080295 and an additional contribution from the National Institute of Neurological Disorders and Stroke. Additional support was provided by National Institute on Aging grant AG023629. Dr. Gardin is a member of the Speakers Bureau for Gilead Sciences. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. A full list of principal Cardiovascular Health Study investigators and institutions can be found at CHS-NHLBI.org.
Presented in part at the American Heart Association Scientific Sessions, November 2009, Orlando, Florida.
- Abbreviations and Acronyms
- body mass index
- body surface area
- coronary heart disease
- cardiovascular disease
- diabetes mellitus
- low-density lipoprotein cholesterol
- left ventricular
- metabolic syndrome
- net reclassification index
- Received February 26, 2015.
- Revision received April 21, 2015.
- Accepted April 23, 2015.
- 2015 American College of Cardiology Foundation
- Ford E.S.
- Malik S.,
- Wong N.D.,
- Franklin S.S.,
- et al.
- Gardin J.M.,
- McClelland R.,
- Kitzman D.,
- et al.
- Burchfiel C.M.,
- Skelton T.N.,
- Andrew M.E.
- Lindman B.R.,
- Arnold S.V.,
- Madrazo J.A.,
- et al.
- Evans J.M.,
- Wang J.,
- Morris A.D.
- Wong N.D.,
- Rozanski A.,
- Gransar H.,
- et al.
- Wong N.D.,
- Sciammarella M.G.,
- Polk D.,
- et al.
- Malik S.,
- Budoff M.J.,
- Katz R.,
- et al.
- Grundy S.M.,
- Cleeman J.I.,
- Daniels S.R.,
- et al.
- SAS Institute
- Kuller L.H.,
- Velentgas P.,
- Barzilay J.,
- et al.
- Jain A.,
- McClelland R.L.,
- Polak J.F.,
- et al.
- Somaratne J.B.,
- Whalley G.A.,
- Poppe K.K.,
- et al.
- Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults
- Hoang K.,
- Ghandehari H.,
- Lopez V.A.,
- Barboza M.G.,
- Wong N.D.
- Bluemke D.A.,
- Kronmal R.A.,
- Lima J.A.,
- et al.