Life-Course Cumulative Burden of Body Mass Index and Blood Pressure on Progression of Left Ventricular Mass and Geometry in Midlife: The Bogalusa Heart Study

Abstract

Rationale:

Data are limited regarding the influence of life-course cumulative burden of increased body mass index (BMI) and elevated blood pressure on the progression of left ventricular (LV) geometric remodeling in midlife.

Objective:

To investigate the dynamic changes in LV mass and LV geometry over 6.4 years during midlife and to examine whether the adverse progression of LV geometric remodeling is influenced by the cumulative burden of BMI and blood pressure from childhood to adulthood.

Methods and Results:

The study consisted of 877 adults (604 whites and 273 blacks; 355 males; mean age=41.4 years at follow-up) who had 5–15 examinations of BMI and blood pressure from childhood and 2 examinations of LV dimensions at baseline and follow-up 6.4 years apart during adulthood. The area under the curve (AUC) was calculated as a measure of long-term burden (total AUC) and trends (incremental AUC) of BMI and systolic blood pressure (SBP). After adjusting for age, race, sex, smoking, alcohol drinking and baseline LV mass index, the annual increase rate of LV mass index was associated with all BMI measures (β=0.16–0.36, P<0.05 for all), adult SBP (β=0.07, P=0.04), and total AUC of SBP (β=0.09, P=0.01), but not with childhood and incremental AUC values of SBP. All BMI and SBP measures (except childhood SBP) were significantly associated with increased risk of incident LV hypertrophy, with odds ratios (ORs) of BMI (OR=1.85–2.74, P<0.05 for all) being significantly greater than those of SBP (OR=1.09–1.34, P<0.05 for all except childhood SBP). In addition, all BMI measures were significantly and positively associated with incident eccentric and concentric LV hypertrophy.

Conclusions:

Life-course cumulative burden of BMI and blood pressure is associated with the development of LVH in midlife, with BMI showing stronger associations than blood pressure.

Graphical Abstract

This community-based longitudinal cohort study followed since childhood demonstrated an adverse progression of LV mass and an increase in the prevalence of LV hypertrophy and abnormal geometric patterns over a 6.4-year period during midlife among white and black adult participants. We found that the annual increase rate of LV mass index was associated with all BMI measures, adult systolic BP and cumulative burden of systolic BP. All BMI and systolic BP measures (except childhood systolic BP) were significantly associated with increased risk of incident LV hypertrophy, with odds ratios of BMI being greater than those of systolic BP. These findings suggest that life-course cumulative burden of BMI and BP is associated with development of LV hypertrophy in midlife, with BMI showing stronger associations than BP.

INTRODUCTION

Both left ventricular (LV) mass (LVM) and LV geometry significantly predict cardiovascular morbidity and mortality.^{1, 2} There is uniformity of view that compared to eccentric hypertrophy (EH), LV concentric hypertrophy (CH) confers a higher risk of cardiovascular outcomes.^{2, 3} Extensive evidence has demonstrated that LV geometric remodeling deteriorates over time, and the progression of LVM and geometry is associated with future cardiovascular events.^{4, 5} It is of importance to identify longitudinal determinants, especially since early life, that lead to adverse evolution of LV geometric remodeling.

Obesity and hypertension, two important cardiovascular risk factors, have been shown to independently contribute to the development of left ventricular hypertrophy (LVH).^6–9 Life-long cohort studies have suggested that risk factors in early life, including high body mass index (BMI) and elevated blood pressure (BP), are associated with adult LVH and geometric remodeling patterns,^10–13 supporting the concept of childhood origins of cardiovascular disease. A growing body of literature indicates that middle age is a critical period for accelerated cardiovascular aging. During the middle age period, silent subclinical alterations in the cardiovascular system begin to transit to clinical manifestations.^{14, 15} To date, data are limited regarding the influence of life-course cumulative burden of BMI and BP on the evolution of cardiovascular structure and changes in LV geometry in midlife.

Utilizing data from the Bogalusa Heart Study with repeated measurements of LV dimensions, we aimed to investigate the dynamic changes in LVM and LV geometry over a 6.4-year period during midlife and to examine whether the adverse progression of LV geometric remodeling is influenced by the cumulative burden of BMI and BP from childhood to adulthood.

METHODS

The data that support the findings of this study are available from the corresponding author on reasonable request.

Study cohort.

The Bogalusa Heart Study, a series of long-term epidemiologic studies in a semirural biracial (65% white and 35% black) community in Bogalusa, Louisiana, was founded by Dr. Gerald Berenson in 1973. This study focuses on the early natural history of cardiovascular disease since childhood.¹⁶ In the community of Bogalusa, Louisiana, 9 cross-sectional surveys of children aged 4 to 19 years and 11 cross-sectional surveys of adults aged 20 to 51 years who were previously examined as children were conducted between 1973 and 2010. Linking these repeated cross-sectional examinations conducted every 2 to 3 years has resulted in serial observations of cardiovascular risk factors from childhood to adulthood in the same individuals. Cardiac ultrasound examinations of LV dimensions in 1182 adults were performed in the baseline survey in 2000–2004. Of these 1182 adults, 898 participants participated in the 2006–2010 follow-up survey. We excluded participants (n=7) who had missing values for any of the study variables in the baseline and/or follow-up surveys and those (n=14) who had ≤4 measurements of BMI and BP (Online Figure I). Finally, the longitudinal cohort of the present study consisted of 877 adult subjects (604 whites and 273 blacks; 355 males; mean age=41.4 years at follow-up). These participants were examined 5–15 times for BMI and BP from childhood to adulthood (9.0 times on average, at least 2 times in childhood and at least 2 times in adulthood). The number of participants who were examined 5, 6, 7 and 8~15 times was 20, 52, 128 and 677, respectively. LV dimensions were measured at baseline and follow-up 6.4 years apart in adulthood. The mean follow-up period from the first measurement (childhood) to the last measurement (adulthood) for BMI and BP was 31.4 years. Baseline characteristics between included and excluded participants are shown in Online Table I. The participants (n=877) who were included in the analysis were about 1 year older than those (n=305) who were excluded from the analysis. Other study variables did not show significant differences between the two groups.

To validate the growth curve modeling, we used BMI and BP data on 3357 participants who were followed 5–18 times (at least 2 times in childhood and at least 2 times in adulthood) in four external longitudinal study cohorts of the International Childhood Cardiovascular Cohort (i3C) Consortium.¹⁷ The four cohorts included the Cardiovascular Risk in Young Finns Study,¹⁸ the Muscatine Study,¹⁹ the NHLBI Growth and Health Study,²⁰ and the Prevention of High Blood Pressure in Children Study. Detailed descriptions of these four individual cohorts are provided in Materials and Methods in the Online Data Supplement. Characteristics of the external validation cohorts are shown in Online Table II.

Written informed consent was obtained from each study participant or from parent/guardian in those younger than18 years old. Study protocols were approved by the Institutional Review Board of the Tulane University Health Sciences Center.

Examinations.

Standardized protocols were used by trained staff members in all surveys since 1973. Height and weight were measured in duplicate, and the mean values were used for analysis. BMI was calculated as weight in kilograms divided by height in meters squared. Systolic blood pressure (SBP) and diastolic blood pressure (DBP) were obtained using a mercury sphygmomanometer on right arms in a relaxed sitting position by 2 trained observers (3 times each) between 8:00 AM and 10:00 AM. The mean values of the 6 readings were used for analysis of BP. For hypertensive (n=178) patients who were under antihypertensive treatment and had SBP/DBP<140/90 mmHg, forced values (140/90 mmHg) were assigned for measured SBP/DBP.

Echocardiographic LV structure measurements.

Identical scanning protocols were used for each echocardiography examination. LV dimensions were assessed by 2-dimensional guided M-mode echocardiography with 2.25- and 3.5-MHz transducers according to American Society of Echocardiography recommendations.²¹ Parasternal long- and short-axis views were collected for measuring LV end-diastolic and end-systolic measurements in duplicate, and the mean was calculated. LVM was calculated from a necropsy-validated formula on the basis of a thick-wall prolate ellipsoidal geometry.²² To take body size into account, LVM was indexed for body height (m^2.7) as LVM index (LVMI). LV relative wall thickness (RWT) was calculated as septal wall thickness plus posterior wall thickness divided by LV end-diastolic diameter.²³ A random sample of 10% of all subjects underwent a repeat echocardiographic examination and was used to estimate measurement errors. The intraclass correlation coefficients between the repeated measurements of LVM were 0.78 at baseline and 0.82 at follow-up.

The presence of LVH was defined by LVMI >46.7 g/m^2.7 in women and >49.2 g/m^2.7 in men; LV geometry was considered concentric when RWT was >0.42.²⁴ Four patterns of LV geometry were defined: 1) normal LV geometry (normal RWT with no LVH), 2) concentric remodeling (CR, increased RWT with no LVH), 3) EH (normal RWT with LVH), and 4) CH (increased RWT with LVH).^23–25 The annual rate of change in LVMI and RWT was calculated as the differences in their values between baseline and follow-up divided by the number of follow-up years. Incident LVH, CR, EH and CH were defined as new cases developed during the follow-up period.

Statistical methods.

Long-term burden and trends of BMI and SBP were measured as the area under the curve (AUC) which was calculated using statistical models we previously described.^26–28 Growth curves of BMI and SBP measured multiple times from childhood to adulthood were constructed using a random-effects mixed model by SAS proc MIXED.^{29, 30} The SAS codes are provided in Materials and Methods in the Online Data Supplement. The mixed model incorporates fixed and random effects and allows the intercept, linear and nonlinear parameters to vary from individual to individual. The random-effect coefficients represent the difference between the predicted values by fixed effect parameters and the observed values for each individual. This model allows for repeated measurements and different numbers of unequally spaced observations across individuals. The unstructured covariance structure or the variance components structure was used based on their Akaike’s Information Criterion (AIC) values.³¹ The model building was based on AIC and p-values of the independent variable (age) at the significance level of 0.05. Age and its higher-order terms were included one by one for the most parsimonious model building. The higher-order terms of age were not included in the model if they were not significant, or made lower-order terms not significant, or did not improve the goodness-of-fit of the model based on AIC values. Quadratic curves were fitted for BMI, and cubic curves for SBP in race-sex groups.

$${\text{BMI}}_{\text{j}}=\left({\mathrm{\beta }}_0+{\text{b}}_{0\text{j}}\right)+\left({\mathrm{\beta }}_1+{\text{b}}_{1\text{j}}\right)\text{ age }+\left({\mathrm{\beta }}_2+{\text{b}}_{2\text{j}}\right){\text{age}}^2+{\mathrm{\epsilon }}_{\text{j}}$$

$${\text{SBP}}_{\text{j}}=\left({\mathrm{\beta }}_0+{\text{b}}_{0\text{j}}\right)+\left({\mathrm{\beta }}_1+{\text{b}}_{1\text{j}}\right)\text{ age }+\left({\mathrm{\beta }}_2+{\text{b}}_{2\text{j}}\right){\text{ age}}^2+\left({\mathrm{\beta }}_3+{\text{b}}_{3\text{j}}\right){\text{ age}}^3+{\mathrm{\epsilon }}_{\text{j}}$$

where β = (β_i)’ is a vector of fixed effect parameters, b = (b_j)’ is a vector of random effect parameters, and ɛ_j is an unknown error term. Age was centered to the mean age (20.1 years) to remove the collinearity of age with its higher-order terms. The term age² was divided by 10, and age³ by 20 to improve the model fitting. Distribution normality of regression residuals of BMI and BP at age points was tested using Shapiro-Wilk test and Q-Q plot after the model building.

We previously applied the same random-effects growth curve modeling methodology to the five longitudinal study cohorts of the i3C Consortium.³² In the current study, we included 3357 participants who were followed 5–18 times since childhood (at least 2 times in childhood and at least 2 times in adulthood) in the other four cohorts for validating the growth curve models.

As shown in Online Figures II and III, the AUCs of BMI and SBP were calculated as the integral of the curve parameters for each participant over the interval from age₁ to age_n, where age₁ was the first childhood age in the baseline survey, and age_n was the last adulthood age in the follow-up survey. Since participants had different follow-up periods, the AUC values were divided by the number of follow-up years. The AUC measures have advantages over other conventional longitudinal analysis models in that they measure both long-term burden and trends. Total AUC (a+b) can be considered a measure of a long-term cumulative burden; incremental AUC (a), determined by within-subject variability, represents a combination of linear and nonlinear longitudinal trends.

The significance of differences between race and sex groups was tested using analysis of covariance and Chi-square test for continuous and categorical variables, respectively. Multivariable linear regression analyses were performed to examine the associations of LVMI at baseline and follow-up and its progression rate with BMI and SBP measures. Multivariable logistic regression analyses were used to examine the associations of incident LVH and geometric patterns with BMI and SBP measures. The covariates included age, race, sex, smoking and alcohol drinking. For association analyses of LVMI progression rate, we also performed the multivariable linear regression models with additional adjustment for baseline LVMI.

Prior to regression analyses, childhood and adulthood values as well as total and incremental AUC values of BMI and SBP were adjusted for corresponding age (or average age) by regression residual analyses and then standardized with Z-transformation (mean=0, SD=1) by race-sex groups in order to avoid collinearity of childhood and adulthood ages in the same model. Also, for analyses of incremental AUC, baseline values of BMI and SBP were included in the model for adjustment to control for the regression-to-the-mean bias. Interaction effects of BMI, BP and race were examined using regression interaction models by including their interaction terms in the model. Bonferroni corrections for multiple tests were used to determine the significance threshold. Multicollinearity among predictor variables and covariates in the models was assessed using correlation matrix (>0.8), tolerance (<0.1), variance inflation factor (>10) and collinearity diagnostics for an eigen-system analysis of covariance comparison in the models.^{33, 34} Statistical power for the total cohort and subgroups was estimated by PASS11 (Power Analysis & Sample Size) software.³⁵

RESULTS

Table 1 summarizes BMI and BP in childhood and adulthood and their AUC values by race and sex. Childhood BMI and BP did not show race and sex differences except for race difference (white>black) in females for DBP. In adulthood, BMI showed sex difference (male<female) in blacks and race difference (white<black) in females. Males versus females and blacks versus whites had higher SBP and DBP. Total and incremental AUC values of BMI, SBP, and DBP showed significant race and sex differences except for race differences in total and incremental AUCs of BMI among males. Pearson correlation coefficients between BMI and SBP were 0.379, 0.322, 0.433, and 0.466 for childhood, adulthood, and total and incremental AUC values, respectively (P < 0.0001 for all).

Table 1.

Body mass index and blood pressure by race and sex

Variable	White			Black			P for  Race Difference
	Male (n=249)	Female  (n=355)	P	Male  (n=106)	Female  (n=167)	P	Male	Female
Childhood (First exam)
Age (yr)	10.3 (3.2)	10.1 (3.4)	0.57	9.8 (2.9)	9.5 (2.8)	0.40	0.17	0.03
BMI (kg/m2)	17.5 (3.4)	17.7 (3.4)	0.42	17.2 (3.4)	17.4 (3.6)	0.58	0.44	0.43
SBP (mmHg)	100.3 (10.2)	99.9 (9.5)	0.90	99.4 (11.0)	98.1 (10.0)	0.37	0.47	0.05
DBP (mmHg)	61.6 (8.2)	62.4 (8.5)	0.18	62.4 (7.5)	60.3 (8.8)	0.05	0.39	0.01
Adulthood (Last exam)
Age (yr)	41.8 (4.5)	41.3 (4.6)	0.17	42.2 (4.2)	40.7 (4.9)	0.01	0.44	0.18
BMI (kg/m2)	29.7 (5.5)	28.5 (7.1)	0.07	30.3 (7.8)	32.7 (8.9)	0.02	0.39	<0.0001
SBP (mmHg)	117.9 (10.9)	111.1 (12.1)	<0.0001	128.4 (18.2)	122.5 (18.1)	0.01	<0.0001	<0.0001
DBP (mmHg)	81.2 (8.0)	76.5 (8.5)	<0.0001	86.8 (12.8)	82.2 (11.2)	0.002	<0.0001	<0.0001
AUC measures
Average age(yr)	25.6 (4.0)	25.3 (4.2)	0.46	24.7 (4.3)	24.1 (3.6)	0.28	0.05	0.002
BMI AUCt	25.1 (4.1)	24.2 (4.9)	0.03	25.2 (5.4)	26.8 (6.0)	0.02	0.96	<0.0001
SBP AUCt	114.1 (7.8)	108.2 (7.0)	<0.0001	118.9 (9.3)	114.0 (9.7)	<0.0001	<0.0001	<0.0001
DBP AUCt	74.1 (5.5)	71.3 (5.0)	<0.0001	76.2 (7.2)	73.7 (6.5)	0.004	0.003	<0.0001
BMI AUCi	7.5 (2.7)	6.6 (3.7)	0.002	7.9 (3.6)	9.2 (4.3)	0.01	0.29	<0.0001
SBP AUCi	13.3 (5.7)	8.4 (5.9)	<0.0001	19.0 (7.6)	15.3 (7.2)	<0.0001	<0.0001	<0.0001
DBP AUCi	12.7 (3.6)	9.3 (4.0)	<0.0001	14.5 (5.1)	12.6 (4.7)	0.001	0.0008	<0.0001

BMI=body mass index; S(D)BP=systolic (diastolic) blood pressure; AUC_t=total area under the curve; AUC_i=incremental area under the curve

Table 2 summarizes the outcome measures in adult surveys by race and sex. In the baseline survey, whites versus blacks had lower LVM, LVMI, and RWT and lower prevalence of LVH and CR only in females, and lower prevalence of CH for both sexes; males versus females had higher LVM for both races. In the follow-up survey, whites versus blacks had lower LVM, LVMI, RWT and prevalence of LVH and CH for both sexes. Males versus females had higher LVM for both races and higher LVMI, RWT and prevalence of CR and EH in whites. Whites versus blacks had lower progression rates of LVM and LVMI only in females; males versus females had greater progression rates of LVM and LVMI in whites.

Table 2.

Left ventricular measurements at baseline and follow-up by race and sex

Variable	White			Black			P for Race Difference
	Male (n=249)	Female (n=355)	P	Male (n=106)	Female (n=167)	P	Male	Female
Adulthood (Baseline)
Age (yr)	35.4 (4.5)	34.7 (4.6)	0.08	35.9 (4.3)	34.5 (4.6)	0.01	0.34	0.52
LVM (g)	155.0 (46.7)	113.5 (34.3)	<0.0001	161.9 (54.4)	124.0 (42.6)	<0.0001	0.22	0.003
LVMI (g/m2.7)	32.8 (9.6)	30.2 (9.1)	0.001	34.8 (11.7)	33.3 (11.2)	0.28	0.09	0.0006
RWT (cm)	0.326 (0.068)	0.317 (0.063)	0.09	0.333 (0.062)	0.333 (0.085)	0.96	0.40	0.014
LVH, n (%)	17 (6.8)	17 (4.8)	0.36	10 (9.4)	19 (11.4)	0.62	0.40	0.006
CR, n (%)	21 (8.4)	15 (4.2)	0.03	7 (6.6)	14 (8.4)	0.57	0.60	0.03
EH, n (%)	16 (6.4)	13 (3.7)	0.14	7 (6.6)	12 (7.2)	0.80	0.94	0.06
CH, n (%)	1 (0.4)	4 (1.1)	0.37	3 (2.8)	7 (4.2)	0.54	0.04	0.01
Adulthood (Follow-up)
Age (yr)	41.8 (4.5)	41.3 (4.6)	0.17	42.2 (4.2)	40.7 (4.9)	0.01	0.44	0.17
LVM (g)	197.2 (51.5)	141.1 (39.7)	<0.0001	212.9 (66.4)	161.3 (50.3)	<0.0001	0.02	<0.0001
LVMI (g/m2.7)	41.7 (10.3)	37.5 (10.5)	<0.0001	45.8 (14.0)	43.0 (13.2)	0.09	0.002	<0.0001
RWT (cm)	0.399 (0.081)	0.385 (0.080)	0.05	0.418 (0.079)	0.401 (0.078)	0.09	0.04	0.03
LVH, n (%)	55 (22.0)	60 (16.9)	0.16	34 (32.1)	58 (34.7)	0.68	0.04	<0.0001
CR, n (%)	67 (26.9)	71 (20.0)	0.01	29 (27.4)	30 (18.0)	0.07	0.38	0.48
EH, n (%)	32 (12.9)	23 (6.5)	0.002	15 (14.2)	28 (16.8)	0.99	0.36	<0.0001
CH, n (%)	23 (9.2)	37 (10.4)	0.92	19 (17.9)	30 (18.0)	0.64	0.01	0.002
Progression rate
Follow-up duration (yr)	6.4 (1.9)	6.5 (1.7)	0.34	6.3 (1.9)	6.2 (2.0)	0.74	0.66	0.05
LVM (g/yr)	6.8 (7.7)	4.3 (5.5)	<0.0001	7.9 (7.3)	6.4 (6.5)	0.07	0.21	<0.0001
LVMI (g/m2.7/yr)	1.4 (1.6)	1.1 (1.4)	0.02	1.7 (1.7)	1.7 (1.7)	0.81	0.14	0.0002
RWT (cm/yr)	0.012 (0.018)	0.011 (0.017)	0.56	0.014 (0.017)	0.012 (0.023)	0.42	0.28	0.29

LVM(I)=left ventricular mass (index); RWT=relative wall thickness; LVH=left ventricular hypertrophy; CR=concentric remodeling; EH=eccentric hypertrophy; CH=concentric hypertrophy

During the follow-up period, among 814 participants with normal LVM at baseline,154 (18.9%) developed LVH, 73 (9.0%) developed EH, and 81 (10.0%) developed CH. Figure 1 shows the incidence rates of LVH and LV geometric patterns by race. Black versus white subjects had higher incidence of LVH (26.6% vs. 15.6%, P<0.0001), EH (13.1% vs. 7.2%, P=0.003) and CH (13.5% vs. 8.4%, P=0.01), but CR showed no race difference (22.3% vs. 20.8%, P=0.28).

Figure 1.

Incidence of LVH and LV geometry patterns by race in the Bogalusa Heart Study cohort

LV(H)=left ventricular (hypertrophy); CR=concentric remodeling; EH=eccentric hypertrophy; CH=concentric hypertrophy

Online Figures IV and V present growth curves of BMI and SBP, respectively, from childhood to adulthood by race-sex group. All curve parameters of BMI and SBP were different from 0 (P<0.0001). The growth curves of BMI in white males and black males had similar trajectories; BMI increased faster in black females than in other race-sex groups, and this difference was discernible around age 10 years. White females tended to have a lower BMI than the remaining race-sex groups after age 20. SBP did not differ markedly during the age range of 4 to 14 years. The growth curves of SBP were separated from about 15 years and beyond, with men having higher levels and slopes than women before 45 years in blacks and before 50 years in whites. Like the models built using the Bogalusa Heart Study cohort, the repeatedly measured BMI and SBP data of 3357 participants in the external cohorts fitted quadratic and cubic curves, respectively (Online Table III).

Figures 2 and 3 show growth curves of BMI and SBP, respectively, in groups of incident LVH and LV geometry patterns. Subjects with incident LVH, EH, and CH had consistently higher levels of BMI and SBP from childhood to adulthood compared to those with persistent normal LVM during follow-up period. In contrast, subjects with incident CR had consistently similar levels of BMI and SBP compared to those with persistent normal RWT during follow-up period. Online Table IV presents detailed information on curve parameters in groups classified by incident LVH and LV geometry patterns. Subjects with incident LVH, EH, and CH, compared to those with normal LVM, had significantly greater values of β₀+b₀ and β₁+b₁ but lower β₂+b₂ for BMI; subjects with incident LVH, EH, and CH, compared to those without incident LVH, had significantly greater values of β₀+b₀, β₁+b₁ and β₂+b₂, but lower values of β₃+b₃ for SBP.

Figure 2.

Growth curves of BMI in subgroups classified by incident LVH and LV geometry patterns in the Bogalusa Heart Study cohort

BMI=body mass index; LV(H)=left ventricular (hypertrophy); CR=concentric remodeling; EH=eccentric hypertrophy; CH=concentric hypertrophy

Shaded areas indicate estimated 95% confidence interval bands.

Figure 3.

Growth curves of SBP in subgroups classified by incident LVH and LV geometry patterns in the Bogalusa Heart Study cohort

SBP=systolic blood pressure; LV(H)=left ventricular (hypertrophy); CR=concentric remodeling; EH=eccentric hypertrophy; CH=concentric hypertrophy

Shaded areas indicate estimated 95% confidence interval bands.

Table 3 shows standardized regression coefficients (βs) of LVMI at baseline and follow-up and its progression rate on BMI and SBP in separate models. After adjusting for adult age, race, sex, smoking, and alcohol drinking, baseline and follow-up LVMI was significantly associated with BMI measures (β=0.29–0.55, P<0.0001 for all) and SBP measures (β=0.06–0.13, P<0.05 for all). The annual increase rate of LVMI was positively and significantly associated with adulthood (β=0.12, P=0.002), total AUC (β=0.09, P=0.01) and incremental AUC (β=0.09, P=0.03) values of BMI but not with childhood BMI (β=0.06, P=0.11); LVMI progression rate was not associated with any SBP measures. With additional adjustment for baseline LVMI, LVMI progression rate was associated with all BMI measures (β=0.16–0.36, P<0.0001 for all), adult SBP (β=0.07, P=0.04), and total AUC of SBP (β=0.09, P=0.01), but not with childhood and incremental AUC values of SBP. The association strengths of BMI measured as βs were consistently and significantly greater than those of SBP for all outcome variables (P<0.05 for difference). After Bonferroni correction for multiple tests, all BMI measures showed significant associations with LVMI progression rate, but SBP measures did not. In addition, all the associations of childhood BMI and SBP with outcomes became nonsignificant after adjusting for adult BMI and SBP (Online Table V).

Table 3.

Standardized regression coefficients of BMI and SBP on LVMI

Independent Variable	Dependent Variable
	Baseline LVMI		Follow-up LVMI		LVMI Progression Rate		LVMI Progression Rate||
	β (SE)	P	β (SE)	P	β (SE)	P	β (SE)	P
Model 1s
Childhood BMI*	0.29 (0.03)	<0.0001	0.31 (0.03)	<0.0001	0.06 (0.04)	0.11	0.16 (0.04)	<0.0001
Childhood SBP*	0.07 (0.03)	0.04	0.08 (0.03)	0.02	0.03 (0.04)	0.43	0.05 (0.04)	0.16
Model 2s
Adulthood BMI†	0.52 (0.03)	<0.0001	0.55 (0.03)	<0.0001	0.12 (0.04)	0.002	0.36 (0.04)	<0.0001
Adulthood SBP†	0.06 (0.03)	0.04	0.10 (0.03)	0.0004	0.05 (0.04)	0.16	0.07 (0.03)	0.04
Model 3s
BMI AUCt‡	0.47 (0.03)	<0.0001	0.50 (0.03)	<0.0001	0.09 (0.04)	0.01	0.31 (0.04)	<0.0001
SBP AUCt‡	0.11 (0.03)	0.001	0.13 (0.03)	<0.0001	0.04 (0.04)	0.31	0.09 (0.03)	0.01
Model 4s
BMI AUCi§	0.40 (0.04)	<0.0001	0.43 (0.03)	<0.0001	0.09 (0.04)	0.03	0.24 (0.04)	<0.0001
SBP AUCi§	0.08 (0.03)	0.02	0.11 (0.03)	0.0007	0.02 (0.04)	0.57	0.05 (0.04)	0.20

BMI=body mass index; SBP=systolic blood pressure; AUC_t=total area under the curve; AUC_i=incremental area under the curve; LVMI=left ventricular mass index; β=standardized regression coefficient; SE=standard error

Model 1 includes childhood BMI and SBP, race, sex, adult age, smoking and alcohol drinking.

Model 2 includes adulthood BMI and SBP, race, sex, adult age, smoking and alcohol drinking.

Model 3 includes total AUC of BMI and SBP, race, sex, adult age, smoking and alcohol drinking.

Model 4 includes incremental AUC of BMI and SBP, race, sex, adult age, smoking and alcohol drinking.

^*, adjusted for childhood age and then Z-transformed (mean=0, SD=1)

^†, adjusted for adulthood age and then Z-transformed (mean=0, SD=1)

^‡, adjusted for average age and then Z-transformed (mean=0, SD=1)

^§, adjusted for average age and baseline values and then Z-transformed (mean=0, SD=1)

^||, additional adjustment for baseline LVMI

P<0.003 was considered significant after Bonferroni correction.

Table 4 shows odds ratios (ORs) of incident LVH and LV geometric patterns associated with BMI and SBP in terms of their childhood, adulthood, total and incremental AUC values in separate models. After adjusting for covariates, higher values of BMI and SBP (except childhood SBP) were significantly associated with increased risk of incident LVH, with ORs of BMI (OR=1.85–2.74, P<0.0001 for all) being significantly greater than those of SBP (OR=1.09–1.34, P<0.05 for all except childhood SBP). In the analysis of LV geometric patterns, all BMI measures were significantly associated with increased risks of incident EH (OR=1.79–2.75, P<0.0001 for all) and CH (OR=1.82–2.90, P<0.0001 for all); however, all SBP values were not associated with incident EH and CH except the association of EH with total and incremental AUCs of SBP. BMI versus SBP had significantly greater ORs for all the outcome variables (P<0.05 for difference). After Bonferroni correction for multiple tests, BMI measures showed significant associations with all outcomes, but SBP measures did not. In addition, all the associations of childhood BMI and SBP with outcomes became nonsignificant after adjusting for adult BMI and SBP (Online Table VI).

Table 4.

Odds ratios of BMI and SBP for incident LVH and LV geometry patterns

Independent Variable	Dependent Variable (Control group: n=660)
	Incident LVH (n=154)		Incident EH (n=73)		Incident CH (n=81)
	OR (95%CI)	P	OR (95%CI)	P	OR (95%CI)	P
Model 1s
Childhood BMI*	1.85 (1.52–2.25)	<0.0001	1.79 (1.39–2.31)	<0.0001	1.82 (1.43–2.31)	<0.0001
Childhood SBP*	1.09 (0.90–1.33)	0.37	1.06 (0.82–1.38)	0.65	1.13 (0.87–1.46)	0.37
Model 2s
Adulthood BMI†	2.74 (2.18–3.43)	<0.0001	2.75 (2.05–3.69)	<0.0001	2.90 (2.18–3.85)	<0.0001
Adulthood SBP†	1.27 (1.03–1.56)	0.02	1.22 (0.92–1.63)	0.16	1.26 (0.97–1.63)	0.09
Model 3s
BMI AUCt‡	2.65 (2.10–3.34)	<0.0001	2.48 (1.87–3.29)	<0.0001	2.67 (2.03–3.53)	<0.0001
SBP AUCt‡	1.34 (1.08–1.67)	0.009	1.38 (1.03–1.84)	0.03	1.32 (1.00–1.74)	0.05
Model 4s
BMI AUCi§	1.94 (1.56–2.41)	<0.0001	1.94 (1.44–2.60)	<0.0001	2.17 (1.64–2.88)	<0.0001
SBP AUCi§	1.33 (1.08–1.63)	0.007	1.36 (1.03–1.81)	0.03	1.26 (0.97–1.65)	0.08

BMI=body mass index; SBP=systolic blood pressure; AUC_t=total area under the curve; AUC_i=incremental area under the curve; LV(H)=left ventricular (hypertrophy); EH=eccentric hypertrophy; CH=concentric hypertrophy; OR=odds ratio; CI=confidence interval

Model 1 includes childhood BMI and SBP, race, sex, adult age, smoking and alcohol drinking.

Model 2 includes adulthood BMI and SBP, race, sex, adult age, smoking and alcohol drinking.

Model 3 includes total AUCs of BMI and SBP, race, sex, adult age, smoking and alcohol drinking.

Model 4 includes incremental AUCs of BMI and SBP, race, sex, adult age, smoking and alcohol drinking.

^*, adjusted for childhood age and then Z-transformed (mean=0, SD=1)

^†, adjusted for adulthood age and then Z-transformed (mean=0, SD=1)

^‡, adjusted for average age and then Z-transformed (mean=0, SD=1)

^§, adjusted for average age and baseline values and then Z-transformed (mean=0, SD=1)

P<0.004 was considered significant after Bonferroni correction.

Online Table VII presents ORs of BMI and SBP measures for incident LVH by race. In whites, incident LVH was significantly associated with BMI measures but not with SBP measures after adjusting for covariates; in blacks, however, incident LVH was associated with both BMI and SBP values except childhood SBP. Of note, blacks versus whites had significantly stronger associations of SBP measures (except childhood SBP) with incident LVH, but no significant race differences in the associations were found for BMI measures.

The interaction effects of BMI with BP in terms of childhood, adulthood, total and incremental AUC values were not significant on baseline and follow-up LVMI, LVMI progression, incident LVH and remodeling patterns (P=0.22–0.87).

DISCUSSION

This community-based longitudinal cohort study demonstrated an adverse progression of LVM and an increase in the prevalence of LVH and abnormal geometric patterns over a 6.4-year period during midlife among white and black adult participants. The annual increase rate of LVMI was associated with all BMI measures, adult SBP and cumulative burden of SBP. All BMI and SBP measures (except childhood SBP) were significantly associated with increased risk of incident LVH, with ORs of BMI being greater than those of SBP. For LV geometric patterns, all BMI measures were positively associated with incident EH and CH; however, only the cumulative burden of SBP was positively associated with incident EH. These findings suggest that life-course cumulative burden of BMI and BP is associated with development of LVH in midlife, with BMI showing stronger associations than BP.

Echocardiography allows for identification of different forms of LV geometric remodeling, including eccentric and concentric hypertrophy and disproportionate septal thickness. Although the significance of the forms is not yet entirely defined, CH is considered to confer a higher risk of cardiovascular events.² Several studies of adults with repeated examinations of LV dimensions have reported that LV geometric remodeling worsens progressively with age, although the progression rates vary by populations.^36–38 The Coronary Artery Risk Development in Young Adults Study showed that the increase in LVM over 5 years was small for all race and sex groups among young adults aged 23–35 years.³⁶ Another publication from the same study with over 20 years of follow-up showed that the prevalence of normal geometry declined from 84.2% to 69.7%, with black women having the most adverse changes.³⁷ The Framingham Heart Study of middle-aged to elderly white adults showed that over a 4-year follow-up period, the progression from normal geometry to CR was common (20%), but the progression to EH (8%) or CH (4%) was lower.⁴ In this study cohort of middle-aged adults, the annual rate of increase in LVM was 5.29 g/year in whites and 6.99 g/year in blacks over a 6.4-year follow-up. The prevalence of LVH increased from 5.6% at baseline to 19.0% at follow-up in whites and from 10.6% at baseline to 33.7% at follow-up in blacks. The variations of the progression rates of LV mass and geometry parameters among different populations suggest the influences of different age, ethnicity, and prevalence of risk factors.

Obesity and hypertension often occur together and place a dual burden on the left ventricle.^6–8 Although the underlying mechanisms are incompletely understood, available evidence has shown that excessive adiposity can affect heart size through chronic volume overload, insulin resistance, and inflammatory response,^{39, 40} and hypertensive LVH results from chronic hemodynamic overload and increased central pressure.^{41, 42} Interventional studies have demonstrated an apparent improvement of LVM and geometry after surgery- or diet-induced weight loss among obese patients⁴³ and BP-lowering therapy among hypertensive patients.^{44, 45} Previous longitudinal studies have demonstrated that higher levels of BMI and BP in childhood are significantly associated with cardiac mass evaluated at one-time point in adulthood.^10–12 The Bogalusa Heart study has previously reported that long-term cumulative burden and trends of BMI and BP from childhood predict increased risks of LVH and LV geometric patterns in adult life.¹¹ However, the impact of life-course cumulative burden of BMI and BP on the progression of LV geometric remodeling has not been reported.

In current study with longitudinal data of BMI and BP from childhood and repeated measurements of LV dimensions during midlife, we expanded previous observations by finding that the life-course cumulative burden of BMI and BP is associated with progression in LVM as well as incident LVH. In addition, BMI showed stronger associations based on comparison of the association parameters, which is supported by previous studies showing that obesity is a stronger determinant of LVH compared to hypertension.^{7, 8, 46} In the LV remodeling pattern analyses, childhood BMI, adulthood BMI, total AUC, and incremental AUC of BMI were all significantly associated with incident EH and CH; however, SBP measures were not associated with incident EH and CH except for the association between total AUC of SBP and incident EH. The lack of significance for the associations of SBP with LV geometric patterns may be due to the small number of participants with EH and CH, and the results need to be confirmed in future studies with a large sample size. Also, by comparing the growth curves of BMI and SBP, the differences in BMI and SBP levels between subjects with and without incident LVH, EH or CH became greater and greater with increasing age. These observations indicate that the effect of excessive body weight and elevated BP on the progression of LVH and geometric remodeling during adulthood is accumulative and exacerbated during the lifetime.

Although the evidence that blacks had greater prevalence of obesity, hypertension and LV hypertrophy than whites is indisputable, data on racial differences in the magnitude of the association of BMI and BP with incident LVH have been scarce. Our previous cross-sectional study showed that BP levels were more strongly associated with LV eccentric and concentric hypertrophy in blacks than in whites.⁴⁷ In the present study, the effect of BMI measures on LV hypertrophy did not differ significantly between blacks and whites; however, adulthood SBP and long-term measures of SBP were more strongly associated with incident LVH in blacks than in whites. These findings support previous studies showing that black hypertensive adults had a greater degree of LVH than white hypertensives.^{48, 49} These observations suggest that blacks might be more susceptible than whites to BP-related adverse changes in the cardiovascular system.

This community-based longitudinal study cohort with repeated measurements of BMI, BP and LV dimensions provides a unique opportunity to examine the impact of life-course cumulative burden of BMI and BP on subclinical alterations in cardiac structure. This study has certain limitations. First, the forced BP values (140/90 mmHg) assigned to hypertensives on pharmacological treatments may result in bias in the growth curve parameter estimation to some extent because these individuals represented a subgroup who, without treatment, would be expected to have the highest BP levels. Second, in the stratified analyses of LVH and remodeling patterns by race, the sample size of subgroups was relatively small with limited statistical power, especially with multiple comparisons. Large-scale population studies are required to validate the findings of this study. Third, more follow-up surveys in the future are needed to examine the impacts of long-term cumulative burden of BMI and BP on the development of heart failure and other clinical events. Fourth, multicollinearity needs to be investigated in depth because co-dependence between BMI and BP is very difficult to tease out.

In summary, the observations of this study indicate that middle-aged white and black adults undergo an adverse progression of LVM and geometry. In addition, the impacts of high levels of BMI and BP on the progression of LVM and LV geometry remodeling begins early in life, and BMI shows greater impacts than BP. These findings emphasize the importance of developing preventive interventions and strategies for controlling excess body weight and elevated BP early in life to reduce the risk of future cardiovascular disease.

NOVELTY AND SIGNIFICANCE.

What Is Known?

• Adverse progression of left ventricular (LV) mass and geometry predicts future cardiovascular events.

• High body mass index (BMI) and elevated blood pressure (BP) in childhood are associated with adult left ventricular hypertrophy and geometric remodeling patterns.

• Data are limited regarding the influence of life-course cumulative burden of BMI and BP on the evolution of LV mass and geometry in midlife.

What New Information Does This Article Contribute?

• We demonstrated an adverse progression of LV mass and geometry over a 6.4-year period during midlife.

• Life-course cumulative burden of BMI and BP is associated with the development of LV hypertrophy in midlife.

• BMI has a stronger influence on progression of LV mass and geometry than BP.

Nonstandard Abbreviations and Acronyms:

AUC

area under the curve

BMI

body mass index

BP

blood pressure

CH

concentric hypertrophy

CR

concentric remodeling

DBP

diastolic blood pressure

EH

eccentric hypertrophy

LV

left ventricular

LVH

left ventricular hypertrophy

LVM

left ventricular mass

LVMI

left ventricular mass index

OR

odds ratio

RWT

relative wall thickness

SBP

systolic blood pressure