Adverse Left Ventricular Remodeling and Age Assessed with Cardiac MR Imaging: The Multi-Ethnic Study of Atherosclerosis

In patients free of cardiovascular disease at baseline, we observed a longitudinal left ventricular (LV) mass increase in men and a slight decrease in women, whereas LV end-diastolic volume decreased and mass-to-volume ratio increased in both men and women.

Abstract

Purpose

To evaluate age-related left ventricular (LV) remodeling during longitudinal observation of a large cohort of asymptomatic individuals who were free of clinical cardiovascular disease at baseline.

Materials and Methods

The applicable institutional review boards approved this study, and all participants gave informed consent. Cardiac magnetic resonance (MR) imaging was used to identify longitudinal changes in LV structure and function in 2935 participants who underwent baseline and follow-up cardiac MR imaging in the Multi-Ethnic Study of Atherosclerosis. Participants were free of clinical cardiovascular disease at baseline. Participants who experienced an incident coronary heart disease event were excluded. Data were analyzed with multivariable mixed-effects regression models in which the outcome was cardiac MR imaging measurement, and the covariates included follow-up time and cardiac risk factors.

Results

Participants were aged 54–94 years at follow-up, and 53% of the participants were women. Median time between baseline and follow-up cardiac MR imaging was 9.4 years. Over this period, LV mass increased in men and decreased slightly in women (8.0 and −1.6 g per decade, respectively; P < .001). In both men and women, LV end-diastolic volume decreased (−9.8 and −13.3 mL per decade, respectively; P < .001), stroke volume decreased (−8.8 and −8.6 mL per decade, respectively; P < .001), and mass-to-volume ratio increased (0.14 and 0.11 g/mL per decade, respectively; P < .001). Change in LV mass was positively associated with systolic blood pressure and body mass index and negatively associated with treated hypertension and high-density lipoprotein cholesterol level. In men, the longitudinal LV mass increase was in contrast to a cross-sectional pattern of LV mass decrease.

Conclusion

As patients age, the LV responds differently in its mass and volume between men and women, although both men and women experience increased concentric LV remodeling with age. In men, the opposition of longitudinal and cross-sectional changes in LV mass highlights the importance of longitudinal study.

^© RSNA, 2015

Online supplemental material is available for this article.

Introduction

Multiple population-based studies (1–4) have shown left ventricular (LV) mass and geometry are independent predictors of cardiovascular disease events. These associations exist in individuals with a prior history of coronary heart disease or heart failure and those without (5,6). Treatment to reduce LV mass has been shown to decrease the incidence of cardiovascular events (7). These epidemiologic and therapeutic relationships support the role of LV mass as an important subclinical marker for cardiovascular disease.

LV mass is associated with multiple sociodemographic and cardiovascular risk factors, including age, sex, body size, smoking history, physical activity, and hypertension (8). Age-related geometric changes in the LV, termed remodeling, are primarily described in cross-sectional studies. Cross-sectional studies have shown both increases (9–12) and decreases (13) in LV mass with age. In a longitudinal Framingham study, LV mass was observed to increase with age (14). Some of the apparent discrepancies in these studies may be related to differences in the methods used to measure LV mass (two-dimensional echocardiography versus magnetic resonance [MR] imaging), differences in cardiovascular disease prevalence at baseline, or both.

We hypothesized that analysis of longitudinal data may help clarify inconsistencies in previous cross-sectional results concerning age-related cardiac remodeling. Thus, the purpose of this study was to evaluate age-related LV remodeling during longitudinal observation of a large cohort of asymptomatic individuals who were free of clinical cardiovascular disease at baseline.

Materials and Methods

Baseline Study Population (MESA Examination 1)

The study population was drawn from a cohort of 6814 individuals who were enrolled in the Multi-Ethnic Study of Atherosclerosis (MESA). Detailed study design and objectives are described elsewhere (15). Briefly, study participants were enrolled between July 2000 and September 2002 from six community-based centers in the United States. At the time of enrollment, participants were aged 45–84 years and had no history of clinical cardiovascular disease, which included coronary artery disease, peripheral vascular disease, cerebrovascular disease, and heart failure. The cohort represents four self-reported racial groups: white, black, Latino, and Asian. The Asian participants were predominantly of Chinese descent. LV structure and function were assessed with cardiac MR imaging in 5004 study participants to determine LV mass, volume, and function by using a fast gradient-echo technique, as described previously (4,16,17). Institutional review boards at each of the six field centers approved the baseline and follow-up study protocols.

Follow-up Examination (MESA Examination 5)

Between April 2010 and February 2012, MESA participants were seen for a fifth follow-up examination that included measurement of height and weight. Questionnaires were administered to assess smoking history, exercise habits, current use of medications, and physician diagnoses of hypertension and diabetes. Brachial artery blood pressure was measured three times in the seated position, and the mean of the second and third measurements was used for analysis. Glucose, triglycerides, total cholesterol, and high-density lipoprotein (HDL) cholesterol levels were measured after 12 hours of fasting. Low-density lipoprotein (LDL) cholesterol levels were calculated with the Friedewald equation (18).

Follow-up MR Imaging (MESA Examination 5)

As part of the fifth follow-up MESA examination, consenting participants underwent cardiac MR imaging at facilities associated with each field center. Imaging was performed with 1.5-T imagers (Magnetom Avanto and Magnetom Espree, Siemens Medical Systems, Erlangen, Germany; Signa HD, GE Healthcare, Waukesha, Wis) with a six-channel anterior phased-array torso coil and corresponding posterior coil elements. The cardiac MR imaging protocol included acquisition of one cine horizontal long-axis section (four-chamber view), at least 12 cine short-axis sections from the atria to the cardiac apex, and one cine vertical long-axis section (two-chamber view). LV dimensions, function, and myocardial mass were assessed with a cine steady-state free precession pulse sequence. Imaging was performed with retrospective electrocardiographic gating for 50 phases (temporal resolution ≤40 msec), 360 × 360 mm field of view, 256 × 192 image matrix, 8-mm section thickness, and 2-mm section gap (for short-axis images).

Cardiac MR imaging data were transferred to a central reading center by using secure transmission software (19). Trained readers were overseen by two cardiac MR imaging physicians (D.A.B., J.A.C.L.), each with over 15 years of clinical and research interpretation experience. The readers measured LV end-diastolic volume (EDV), LV end-systolic volume (ESV), and LV end-diastolic mass with the CIM software package (version 6.2; Auckland MR Imaging Research Group, University of Auckland, Auckland, New Zealand) (20). Papillary muscles were included in the LV volumes and excluded from the LV mass. Readers had no knowledge of participant risk factors or other clinical information. LV stroke volume was calculated as LV EDV minus LV ESV, and ejection fraction (EF) was calculated as follows: EF = [(LV EDV–LV ESV)/LV EDV]·100.

For quality control purposes, all readers independently analyzed every 10th consecutive cardiac MR imaging study. The overall interobserver intraclass correlation coefficients for LV mass and LV EDV were 0.95 and 0.96, respectively, and technical errors of measurement (21) were 6.1% and 5.4%, respectively. Interobserver agreement was similar to that in the baseline study (16).

The follow-up imaging protocol used an MR imaging pulse sequence, steady-state free precession, that is currently the standard of care, but this sequence was not available during the baseline examination. Steady-state free precession enables a faster examination of higher quality (22), but it is known to produce LV mass measurements that are smaller than those obtained with the fast gradient-echo pulse sequence of the baseline examination (23). To adjust for this difference, calibration curves were applied to each participant’s baseline LV mass and volume measurements. These curves were obtained from 498 randomly chosen participants who underwent fast gradient-echo MR imaging in addition to steady-state free precession MR imaging during the follow-up examination (Appendix E1 [online]). Besides the pulse sequence differences, these calibration curves also enabled adjustment for potential differences associated with reader, analysis software, and scanner equipment changes between the baseline and follow-up examinations. All calibration curves were found to be linear and were fitted with ordinary regression methods.

Statistical Analysis

Data were analyzed with two sets of multivariable mixed-effects regression models. Each of the cardiac MR imaging parameters was the outcome (dependent) variable of one model in each set. Model 1 describes age and birth cohort effects, with sex as a potential effect modifier. Thus, age at baseline, follow-up time, and sex were the covariates (independent variables) for model 1, with interaction terms between sex and the other two covariates. An interaction term between all three covariates represented how longitudinal changes varied by age. The time covariate was equal to 0 for the baseline cardiac parameters. For the follow-up parameters, the time variable was equal to the elapsed time between baseline and follow-up examinations. Time was treated as a random effect with random slopes and intercepts among participants (Appendix E1 [online]).

Model 2 added key traditional cardiovascular risk factors, similar to those appearing in published baseline models (13,17), to describe how the longitudinal change of each cardiac MR imaging parameter was affected by each risk factor. Model 2 also contained interaction terms between follow-up time and all other covariates. Each interaction term represented the corresponding covariate effect on the longitudinal change of the outcome variable, while the linear terms adjusted for baseline values (Appendix E1 [online]).

The risk factors in model 2 were considered fixed effects that were constant between the baseline and follow-up time points. Model 2 covariates determined from the follow-up MESA examination were age, sex, race or ethnicity (four categories), current use of antihypertensive medications (yes or no), current use of lipid-lowering medications (yes or no), current smoking status (yes or no), and glucose metabolism status. Glucose metabolism was represented by three mutually exclusive categories: diabetes (fasting glucose level ≥126 mg/dL [≥7.0 mmol/L] or use of insulin or hypoglycemic medication), impaired fasting glucose (not diabetic with fasting glucose level ≥100 mg/dL [≥5.6 mmol/L] and <126 mg/dL [<7.0 mmol/L]), or normoglycemic (not diabetic and fasting glucose level <100 mg/dL [<5.6 mmol/L]). Means of two measurements, baseline and follow-up, were used as covariates for the following continuously distributed variables: systolic blood pressure, diastolic blood pressure, body mass index (BMI), LDL cholesterol, and HDL cholesterol.

The models’ fit to the data was assessed by inspecting plots of residuals against all covariates and against predicted values to look for nonrandom patterns that may indicate nonlinearity or heteroscedasticity. All analyses were conducted with the Stata software package (version 13.1; Stata, College Station, Tex) and were based on the MESA data set released in June 2013. P < .05 indicated a significant difference.

Results

Of the 5004 MESA participants who underwent baseline cardiac MR imaging, 3565 (71%) returned for the follow-up MESA examination. Of these 3565 participants, 2981 (84%) were eligible for and consented to undergo follow-up cardiac MR imaging. Of these 2981 participants who underwent both baseline and follow-up cardiac MR imaging, we excluded 46 participants who experienced actual coronary heart disease events, defined as myocardial infarction or resuscitated cardiac arrest. The remaining 2935 participants who underwent follow-up were the subjects of analysis in this study.

The median time between baseline and follow-up cardiac MR imaging was 9.4 years. When compared with the 5004 baseline participants, the participants who underwent follow-up MR imaging were younger and had a higher level of education, on average; a higher proportion of white participants and a lower proportion of Latino participants underwent follow-up MR imaging (Table 1). In addition, among participants who underwent follow-up MR imaging, there was lower systolic blood pressure; fewer participants with hypertension, smoking history, or diabetes; and a slightly greater amount of exercise when compared with those who underwent only baseline MR imaging (Table 1).

Table 1.

Comparison of Baseline Characteristics in Baseline Participants versus Follow-up Cardiovascular MR Imaging Participants

Note.—Data in parentheses are raw data used to calculate percentages. Unless otherwise indicated, data are mean ± standard deviation. The denominators vary because of missing data. To convert HDL and LDL cholesterol to SI units (millimoles per liter), multiply by 0.0259.

*P values were calculated with t or χ² tests to compare baseline values in the participants who underwent cardiovascular MR imaging follow-up (n = 2935) with those in participants who did not undergo follow-up imaging (n = 2069, group data not shown).

Among the 2935 participants, mean age was 69.0 years ± 9.2 (standard deviation) (range, 54–94 years), with 53.0% of the participants being women; 42.3%, white; 24.9%, black; 20.3%, Latino; and 12.5%, Asian. At the time of follow-up, 56.5% of participants had hypertension, 20.7% had impaired fasting glucose, 16.9% had diabetes, 7.5% were smokers, and 36.8% were taking lipid-lowering medication.

Sex-specific mean cardiac MR imaging measurements for the follow-up examination are listed in Table 2 along with the mean baseline values. Longitudinal changes observed in LV mass and volume are shown along with their cross-sectional patterns in the Figure. Coefficients from model 1 describe these longitudinal and cross-sectional patterns quantitatively (Table 3). In the Figure, individual line segments represent observed longitudinal changes for each age group, and the trends between successive line segments represent cross-sectional patterns among the age groups. If period effects are assumed to be constant, then each individual line segment in the Figure represents an isolated age-related effect, and the trends between successive line segments for each sex represent a combination of age and birth cohort effects (24). For men, LV mass increased longitudinally (8.0 g per decade from model 1) for all age groups; this was in contrast to a decrease (−3.7 g per decade) when LV mass was viewed cross-sectionally. For women, LV mass decreased slightly (−1.6 g per decade) when viewed longitudinally and did not change when viewed cross-sectionally. In both men and women, LV EDV decreased, stroke volume decreased, and mass-to-volume ratio increased when the LV mass was viewed either longitudinally or cross-sectionally. All of these relationships were significant in model 1 (Table 3). When stratified by race or ethnicity, longitudinal and cross-sectional relationships within strata were found to be qualitatively consistent with the overall relationships.

Table 2.

Sex-specific LV Mass and Volume in Participants Who Underwent Both Baseline and Follow-up Cardiovascular MR Imaging

Note.—Unless otherwise indicated, data are mean ± standard deviation.

*P values from paired t tests to compare follow-up values with baseline values.

^†Indexed by body surface area.

Figure a:

Age-related longitudinal and cross-sectional changes of LV mass and volume by sex. Individual line segments represent observed longitudinal changes for each age group, and the trends between successive line segments represent cross-sectional patterns among the age groups. (a) LV mass increased longitudinally and decreased cross-sectionally in men. (b) LV EDV and (c) stroke volume decreased both longitudinally and cross-sectionally in both men and women. (d) Mass-to-volume ratio increased both longitudinally and cross-sectionally in both men and women.

Table 3.

Sex-specific Cross-sectional and Longitudinal Differences in LV Mass and Volume (Model 1) in Participants Who Underwent Both Baseline and Follow-up Cardiovascular MR Imaging

Note.—Data in parentheses are 95% confidence intervals.

*P < .001.

^†Represents how the longitudinal change changes with baseline age.

^‡P < .05.

^§P < .01.

Figure b:

Age-related longitudinal and cross-sectional changes of LV mass and volume by sex. Individual line segments represent observed longitudinal changes for each age group, and the trends between successive line segments represent cross-sectional patterns among the age groups. (a) LV mass increased longitudinally and decreased cross-sectionally in men. (b) LV EDV and (c) stroke volume decreased both longitudinally and cross-sectionally in both men and women. (d) Mass-to-volume ratio increased both longitudinally and cross-sectionally in both men and women.

Figure c:

Age-related longitudinal and cross-sectional changes of LV mass and volume by sex. Individual line segments represent observed longitudinal changes for each age group, and the trends between successive line segments represent cross-sectional patterns among the age groups. (a) LV mass increased longitudinally and decreased cross-sectionally in men. (b) LV EDV and (c) stroke volume decreased both longitudinally and cross-sectionally in both men and women. (d) Mass-to-volume ratio increased both longitudinally and cross-sectionally in both men and women.

Figure d:

Age-related longitudinal and cross-sectional changes of LV mass and volume by sex. Individual line segments represent observed longitudinal changes for each age group, and the trends between successive line segments represent cross-sectional patterns among the age groups. (a) LV mass increased longitudinally and decreased cross-sectionally in men. (b) LV EDV and (c) stroke volume decreased both longitudinally and cross-sectionally in both men and women. (d) Mass-to-volume ratio increased both longitudinally and cross-sectionally in both men and women.

The three-way interaction term of model 1 (last two rows of Table 3) shows that longitudinal changes in LV mass were slightly less positive (or more negative) in older age groups for both men and women. In women, longitudinal LV EDV reduction was slightly greater in older age groups. These interactions are seen as a slight lack of parallelism among line segments within the corresponding bands in the Figure.

Table 4 lists selected terms from model 2. These selected terms are the model’s interaction terms between risk factors and elapsed time; thus, they indicate how the risk factors affect the longitudinal change of LV mass and volume in a paired comparison of two time points (baseline and follow-up) after adjustment for sociodemographic and baseline risk factors. Systolic blood pressure was positively associated with longitudinal change in LV mass, LV EDV, and stroke volume (P ≤ .002 for all). BMI and use of antihypertensive medication were positively and negatively associated with change in LV mass, respectively (P ≤ .001 for all). Diastolic blood pressure, smoking history, and BMI were negatively associated with changes in both LV EDV and stroke volume (P < .001 for all).

Table 4.

Changes in LV Mass and Volume in Relation to Traditional Cardiovascular Risk Factors (Model 2) in Participants Who Underwent Both Baseline and Follow-up Cardiovascular MR Imaging

Note.—Data in parentheses are 95% confidence intervals. BP = blood pressure, SD = standard deviation.

*Selected interaction terms from model 2, adjusted for sex and race.

^†P < .005.

^‡P < .05.

Residual plots for LV mass and volume from both models 1 and 2 showed no nonrandom patterns that would suggest nonlinearity or substantial heteroscedasticity.

Discussion

In this longitudinal study of a cohort free of clinical cardiovascular disease at baseline, we observed a significant age-related longitudinal increase in LV mass in men and a slight longitudinal decrease in women. We observed significant longitudinal decreases in LV EDV and stroke volume in both men and women. These changes combined to produce significant increases in adverse remodeling of the LV, as shown by increasing mass-to-volume ratios in both men and women.

In contrast to our findings, longitudinal data from the Framingham study, in which two-dimensional echocardiography was used with 16-year follow-up, showed that LV mass increased sharply over time in both men and women, with a greater increase observed in women than in men (14). The apparent discrepancy may be related to clinical differences in the cohorts or to differences in how LV mass was determined between the two studies (cardiac MR imaging versus two-dimensional echocardiography). More recently, cross-sectional data from three-dimensional echocardiography revealed no significant change between LV mass and age in either men or women (25). In the same cross-sectional study (25), LV EDV decreased with age so that mass-to-volume ratio was observed to increase with age in both men and women, similar to findings in the present study. It is possible that the geometric assumptions required by two-dimensional echocardiography to determine LV mass result in an imaging technique that is more correlated with LV mass-to-volume ratio than true LV mass when compared with three-dimensional imaging techniques, such as cardiac MR imaging and three-dimensional echocardiography.

The longitudinal decrease in LV EDV and stroke volume we observed in participants who underwent follow-up is consistent with cross-sectional age relationships reported in the MESA cohort at baseline (13). More recent cross-sectional studies in healthy volunteers with three-dimensional echocardiography have shown an inverse relationship between LV volume and age (25–27). These volume relationships are in contrast to the findings of earlier studies that reported LV volume increased with age (28,29). However, those earlier studies used radionuclide ventriculography, which is an indirect measurement of LV volume that relies on a number of image processing corrections and geometric assumptions that cardiac MR imaging does not require (30).

This study provided an opportunity to compare longitudinal results with cross-sectional results in the same group of patients. We found that LV mass and volume had similar sex-specific longitudinal and cross-sectional trends except for one measurement group: LV mass in men (Figure). The LV mass in men implies an underlying cohort effect that operates in a direction opposite of the age effect. This apparent cohort effect may reflect a true birth cohort effect; however, another strong possibility is that these results may be the consequence of excluding participants with clinical cardiovascular disease at baseline, producing a study population of healthier individuals. Thus, earlier cohorts (represented by line segments toward the right side of the graphs in the Figure, part a) were physiologically younger; therefore, they had smaller LV masses at baseline. Regardless of baseline LV mass, we observed that longitudinal increases in LV mass in men were similar across all age groups, as indicated by the similar slopes of the individual line segments in the Figure (part a). We also observed that for both men and women, the longitudinal effect (age alone) of increasing mass-to-volume ratio was steeper than the cross-sectional effect (age and birth cohort).

The multivariable risk factor models (model 2) were dominated by variables related to blood pressure, BMI, and smoking history. It is not surprising that blood pressure is one of the dominant variables because it has been found to be the most important predictor of LV hypertrophy in many population studies (8). The pattern of positive correlation between systolic blood pressure and LV mass and volume and negative correlation between diastolic blood pressure and LV volume has also been observed in a number of population studies (8).

In both cross-sectional and longitudinal studies, obesity has been found to be most commonly associated with a pattern of concentric LV remodeling (increased mass, decreased volume). We observed similarly in this study that obesity, as indicated by BMI, was associated with a longitudinal increase in LV mass and a decrease in LV volume. We also found that smoking was associated with a longitudinal decrease in LV volume. A possible explanation for this relationship is the observation that smoking may be associated with myocardial fibrosis, which affects the volume of the LV cavity (31,32). Smoking is also associated with LV dysfunction in the MESA cohort (33).

In patients with hypertension, cardiac myocyte sarcomeres are added in parallel, causing myocardial wall thickness to increase out of proportion to the volume of the LV cavity (34). Thus, a relative decrease in LV volume might be explained by an increase in LV mass. However, we observed that in women, LV volume decreased with age even though there was no corresponding increase in LV mass. This finding suggests that some other mechanism may be diminishing LV volume in women or that the causal relationship between mass and volume in women may be different than that in men. One possible hypothesis is that these results reflect sex-specific variations in diffuse myocardial fibrosis, which may inhibit LV volume changes during remodeling. In the MESA cohort, there is evidence from myocardial T1 mapping that diffuse myocardial fibrosis is greater in women but is associated with a greater age-related increase in men (35). The sex-specific LV mass and volume differences observed in the present study may also signal the need for sex-specific treatment approaches for the adverse effects of myocardial remodeling, such as heart failure.

Our study was subject to some limitations. Of the 5004 MESA participants who underwent baseline cardiac MR imaging, follow-up MR imaging data were obtained for only 60% (n = 2981). Because of response and survivor bias, the follow-up study population was disproportionately healthier than the baseline population. Motivation to adopt healthier behaviors also may have occurred. These factors are important sources of potential selection bias. However, we do not believe selection bias would alter the longitudinal results qualitatively. If selection bias was absent and if the study population was less healthy, we would expect to observe directionally similar age-related longitudinal changes to an even greater degree than what we observed in the present study.

Multiple imputation has been used to adjust for missing data, so we explored the possibility of using this statistical technique to adjust for the potential selection bias associated with incomplete follow-up in our study. However, the incomplete follow-up caused missingness of the cardiac MR imaging measurements, which were the main outcome variables. Unfortunately, multiple imputation has been shown to provide no substantial advantage when the missingness involves the outcome variable (36), as in our study.

When interpreting results from the MESA study, it should be noted that the MESA cohort was not intended to be a population of completely healthy individuals. The goal of MESA was to examine correlates and progression of subclinical cardiovascular disease, so by design the MESA cohort included participants with conditions that are known risk factors for cardiovascular disease, such as diabetes and hypertension. While MESA participants may be considered to be relatively healthy since patients with overt cardiovascular disease were excluded at baseline, the study population still represents a spectrum of cardiovascular disease risk.

Individuals in epidemiologic studies are subject to three sources of temporal effects: age, cohort, and period. Because these three temporal effects are collinear, it is mathematically impossible to estimate all three effects simultaneously from a data set without applying an additional constraint (24). In interpreting the results of this study, we assumed that period effects were constant, meaning that factors affecting all individuals equally, such as environment and medical practice, were assumed to be constant during the study period, on average. Of the three possible temporal effects, potential period effects influence the shortest length of time, which is the 10-year study period. Both age and cohort effects operate over a participant’s entire lifetime, which is much longer than the study period. Furthermore, both age and cohort are associated with well-known cardiovascular risk factors. Past trends, such as those involving obesity, diet, and exercise, are major cohort effects since they have a differential influence on successive cohorts.

In summary, in a cohort of individuals who were free of clinical cardiovascular disease at baseline, we observed a longitudinal LV mass increase in men and a slight decrease in women, while LV EDV decreased and mass-to-volume ratio increased longitudinally for both men and women. These longitudinal trends are congruent with previous cross-sectional observations in this cohort. A notable exception to this congruence was observed in men, where the longitudinal increase in LV mass was opposite to the cross-sectional decrease with age that was observed previously. These results highlight the importance of longitudinal study and suggest sex-specific differences in age-related cardiac remodeling. Further research is needed to understand the mechanism behind these sex-specific differences.

Advances in Knowledge

• ■ The aging left ventricle (LV) responded differently in terms of mass and volume between men and women.

• ■ LV mass increased in men (8.0 g per decade, P < .001) and decreased slightly in women (−1.6 g per decade, P < .001) after nearly a decade of longitudinal follow-up; however, mass-to-volume ratio increased similarly in both men and women (0.14 and 0.11 g/mL per decade, respectively; P < .001).

• ■ In men, the longitudinal changes in LV mass were in opposition to cross-sectional changes observed in past studies, highlighting the importance of the longitudinal study.

Implication for Patient Care

• ■ The study results shed light on the pathophysiology of heart failure in elderly patients, particularly heart failure with preserved ejection fraction.