Changes in Truncal Obesity and Fat Distribution Predict Arterial Health

Abstract

Background:

Truncal obesity is associated with metabolic syndrome and cardiovascular risk. Although vascular health is influenced by weight, it is not known whether changes in fat distribution modulate arterial function.

Objective:

We assessed how changes in truncal (android) fat at one year affect arterial stiffness and endothelial function.

Methods:

We recruited 711 healthy volunteers (235 males, age 48±11) into the Emory Predictive Health Study; 498 returned at one year. Measurements included anthropometric and chemistry panels, fat mass using dual-energy X-ray absorptiometry, arterial stiffness indices [pulse wave velocity (PWV), augmentation index (AIx), and subendocardial viability ratio (SEVR)] (Sphygmocor), flow-mediated dilation (FMD) and reactive hyperemia index (RHI, Endo-PAT).

Results:

At baseline, measures of body mass correlated with PWV, AIx, SEVR, and FMD. A multivariable analysis including body mass index (BMI) and traditional risk factors, BMI remained an independent predictor of PWV, AIx, SEVR, and FMD. In a model including BMI and measures of fat distribution, android fat remained an independent predictor of PWV (β= .31, p= .004), AIx (β = .24, p= .008), and SEVR (β = −.41, p <.001). The one-year change in android fat correlated negatively with change in SEVR (β = −.13, p= .005) and FMD (β = −.13, p= .006) after adjustment for change in gynoid fat.

Conclusion:

In addition to BMI, android fat is a determinant of arterial stiffness, independent of traditional risk factors. Changes in android fat over time are associated with simultaneous changes in vascular function, indicating fat distribution’s effect on vascular health.

Introduction:

Excessive abdominal or visceral fat, known as truncal obesity, is associated with metabolic syndrome and cardiovascular risk.^1–5 Obesity and the metabolic syndrome contributes to a systemic pro-inflammatory and oxidized milieu leading to arterial stiffness and endothelial dysfunction. Adipokines are associated with upregulation of pro-fibrotic factors and promote the synthesis of proteins in the extra-cellular matrix of the vessel wall.^6–8 Furthermore, dysfunctional perivascular adipose tissue is related to a local pro-inflammatory state.^{9, 10} Recent data suggests that changes in visceral fat may influence cardiovascular risk factors.¹¹ Specifically, increasing visceral fat over time has been associated with a higher likelihood of hypertension, hyperlipidemia, and metabolic syndrome.¹² Vascular endothelial dysfunction is a precursor for the development of atherosclerosis and predicts incident risk of cardiovascular morbidity and mortality.^{13, 14} Although arterial health is influenced by weight, it is not known whether changes in fat distribution modulate arterial function. In our study, we hypothesized that changes in truncal (android) fat as measured by dual-energy x-ray absorptiometry (DXA) would predict changes in arterial stiffness and endothelial dysfunction.

Methods:

711 healthy volunteers (469 females and 235 males, mean age 48 years ± 11 years) were recruited as part of the Predictive Health Institute from 12/2007–12/2010 at the Emory-Georgia Tech Center for Health Discovery and Well Being (CHDWB). At the baseline visit, each subject was assigned a health partner trained to utilize subjects’ data profiles and collaboratively generate a health goal and personalized action plan at each visit. Participants with acute illness, cerebrovascular disease, heart failure, and coronary or valvular heart disease were excluded. Vascular testing and blood draws were performed after an overnight fast. 498 subjects returned at one year for repeat testing. The study was approved by the Emory University Institutional Review Board and informed consent was obtained from all subjects. Our analysis is a substudy of the overall Emory Predictive Health Initiative focused on the evaluation of cardiovascular risk factors and risk factor reduction in a relatively healthy population.

Hypertension, hypercholesterolemia, and diabetes mellitus were self-reported by questionnaire on each examination. Medication lists were reviewed and subjects were reclassified as having hypertension, hypercholesterolemia, or diabetes mellitus if the medication list reflected medications treating these conditions. Tobacco use was self-reported by questionnaire on each examination. Blood pressure was measured in the seated position 3 times at 5-minute intervals by an automatic device (Omron, Kyoto, Japan) and documented as the mean value. Fasting lipid profile, metabolic panel, and high-sensitivity C-reactive protein were measured at each visit (Quest Diagnostics, Madison, New Jersey, USA). The ten-year risk of coronary death or nonfatal myocardial infarction was estimated by the Atherosclerotic Cardiovascular Disease in Adults (ASCVD) pooled cohort equations.¹⁵

Body Mass Index (BMI) was calculated as weight in kilograms/(height in meters)². Waist and hip circumferences were measured in centimeters (cm) and recorded as the mean of two measurements. The waist:hip ratio was calculated by dividing the mean waist circumference by the mean hip circumference at each visit. Body composition variables were calculated by dual-energy x-ray absorptiometry (GE Lunar Densitometry, iDXA®). DXA is considered a gold standard for body composition analysis and can identify whole-body fat mass within 2% coefficient of variation.¹⁶ The android region included an area from the top of the iliac crest to 20% of the distance from the iliac crest to the bottom of the subject’s head. The gynoid region extended from the top of the greater trochanter down a distance twice the height of the android region.¹⁷ Measures of android and gynoid fat were reported as fat mass in kilograms (kg) from which the android:gynoid fat ratio was derived.

Endothelium-dependent brachial arterial flow-mediated dilation (FMD) is a non-invasive measure of conductance vessel endothelial function and nitric oxide bioavailability and its measurement has been previously described.^{18, 19} In our laboratory, the mean difference in FMD between assessments performed in 11 subjects on consecutive days was 1.26% (SD 0.76), with a correlation coefficient of 0.75. The mean difference in the FMD between 2 readings of the same 11 measurements was 0.82% (SD 0.48, r = 0.97).

Pulse wave velocity (PWV), the central augmentation index (AIx), and the subendocardial viability ratio (SEVR) are measured by application tonometry and provide non-invasive indices of arterial stiffness, arterial wave reflections, and myocardial workload and perfusion respectively. Both PWV and AIx are independent predictors of adverse cardiovascular outcomes.²⁰ SEVR has been associated with coronary artery disease and decreased coronary flow reserve.²¹ PWV, AIx, and SEVR were estimated by the SphygmoCor device (AtCor Medical, Australia), which records pressure waveforms peripherally using a high-fidelity tonometer.^{19, 22, 23} AIx was standardized to a heart rate of 75 beats/min and divided by the subject’s height yielding a height and heart rate corrected value (AIx). Reproducibility studies in our laboratory on 9 subjects on consecutive days have demonstrated a coefficient of variation of 3.8%, 13.8%, and 20.3% for PWV, SEVR, and AIx, respectively.²²

The reactive hyperemia index (RHI) provides an index of post-ischemic microvascular vasodilation that is reflective of endothelial function. Low RHI is associated with adverse cardiovascular events.²⁴ Digital pulse amplitude tonometry is a plethysmographic device allowing detection of pulsatile arterial volume changes by a pressure transducer (Endo-PAT, Itamar-Medical, Israel).²⁵ Full details of the probe technology and derivation of RHI have been described previously.²⁶

Study variables are described as the mean ± SD (unless otherwise specified) for continuous variables or as counts or proportions for categorical variables. Group differences were evaluated by Student t-tests and proportional differences by two-proportion z-tests. Multivariate linear regression models were constructed to determine relationships between measures of fat distribution and measures of endothelial function and arterial stiffness controlling for age, gender, ethnicity, tobacco use, diabetes, mean arterial pressure, current antihypertensive or statin use, total cholesterol, HDL-c, and high-sensitivity C-reactive protein. Beta-coefficients are reported in the multivariate linear regression models to describe the independent variable’s unit change effect on the dependent variable. For example, if an independent variable has a beta-coefficient of 0.25, a unit increase of 1 in the independent variable would increase the dependent by 0.25. For comparison between models, standardized beta-coefficients are reported.

At one-year follow-up, the first difference of the one-year and baseline values of measures of fat distribution, endothelial function and arterial stiffness were used to examine change over time. Bivariate and multivariate analyses of the first difference values were performed to determine relationships between the change in measures of fat distribution and change in measures of endothelial function and arterial stiffness. The “Δ” symbol indicates a first-difference change value. Multivariate analyses of one-year change were controlled for gender, ethnicity, tobacco use at baseline, baseline age, baseline hypertension, baseline diabetes, change in tobacco use at one year, change in HDL, change in total cholesterol, change in gynoid fat mass, and new statin, antihypertensive, or diabetic medication use after one year. As many of the non-invasive measures of arterial stiffness are associated with blood pressure, change in mean arterial pressure (MAP) is also included for comparison. Statistical analyses were conducted with Statistical Package for Social Sciences (IBM SPSS 23, Inc., Chicago, IL, USA).

Results:

Of the 711 subjects, 66% were female, 72% were Caucasian, and the mean age was 48 ± 11 years. Nearly one-third of subjects reported a history of hypertension (Table 1). The mean BMI was 27.8 kg/m² (Table 1). Baseline android and gynoid fat mass were highly correlated with hypertension, diabetes mellitus, blood pressure, and dyslipidemia as well as BMI and waist circumference. Android fat mass positively correlated and gynoid fat negatively correlated with waist:hip ratio (Supplemental Table 2). All fat measures were positively correlated with PWV indicating increased arterial stiffness with higher body mass. BMI, waist circumference, gynoid fat mass, and android fat mass were negatively correlated with SEVR. All weight measures were negatively correlated with FMD except for gynoid fat which was positively correlated. Increased BMI, waist circumference, waist:hip ratio, android fat mass, and android:gynoid fat ratio correlated with RHI (Supplemental Table 3).

Table 1:

Baseline Characteristics with paired T-test of 1-year change

Table 1	Baseline Characteristics (n=711) (Mean ± SD or %)	Baseline follow-up cohort (n=498) (Mean ± SD or %)	One-year follow-up (n=498) (Mean ± SD or %)	p-value
Age (years)	48 ± 11	49 ± 11	50 ± 11	<0.001
Gender (% female)	66%	64%	64%	---
Caucasian	72%	73%	73%	---
Hypertension*	34%	34%	36%	0.004
Hyperlipidemia†	16%	18%	21%	0.010
Current Tobacco Use	6%	6%	4%	0.011
Diabetes‡	11%	11%	13%	0.007
Systolic BP (mmHg)	121 ± 16	121 ± 16	116 ± 15	<0.001
Diastolic BP (mmHg)	76 ± 11	77 ± 15	74 ± 11	<0.001
Total cholesterol (mmol/L)	5.02 ± 0.93	5.02 ± 0.93	4.86 ± 0.91	<0.001
LDL-c (mmol/L)	2.84 ± 0.83	2.87 ± 0.83	2.72 ± 0.78	<0.001
Weight (kg)	79.5 ± 19.7	79.1 ± 19.5	77.8 ± 19.1	<0.001
BMI (kg/m2)	27.8 ± 6.4	27.6 ± 6.1	27.1 ± 5.9	<0.001
Android Fat Mass (kg)	2.76 ± 1.52	2.73 ± 1.49	2.63 ± 1.45	<0.001
Gynoid Fat Mass (kg)	5.66 ± 2.36	5.57 ± 2.27	5.44 ± 2.26	<0.001
RHI	2.15 ± 0.62	2.18 ± 0.65	2.06 ± 0.57	<0.001
PWV (m/s)	7.25 ± 1.32	7.30 ± 1.33	7.05 ± 1.23	<0.001
AIx (height adjusted)	3.83 ± 1.98	3.93 ± 1.93	3.74 ± 1.91	<0.001
SEVR	166 ± 31.2	166 ± 29.9	165 ± 29.4	0.292
Allometrically-scaled FMD (%)	6.65 ± 1.49	6.65 ± 1.48	6.58 ± 1.43	<0.001

Baseline characteristics and T-tests at follow-up. The second column “baseline characteristics” includes the mean ± standard deviation or percentage of the corresponding variable. The third and fourth columns represent a paired T-test of change at one-year. Only those individuals that returned for one-year follow-up were included in the paired T-test (thus, the difference in N between column two and three).

Notes:

^*Hypertension – reported on initial questionnaire or currently using antihypertensives

^†Hyperlipidemia – reported on initial questionnaire or currently using lipid lowering medication

^‡Diabetes – reported on initial questionnaire or currently using diabetic medication.

Multivariate analyses were performed after adjustment for traditional cardiovascular risk factors. These showed that BMI was an independent negative predictor of AIx, SEVR, and FMD, and a positive predictor of PWV (Table 2). Thus, higher BMI is independently associated with increased arterial stiffness and wave reflections (PWV, AIx, SEVR) and reduced endothelium-dependent vasodilation (FMD). After including android fat in the model, android fat remained an independent positive predictor of PWV and a negative predictor of SEVR (Table 4). Even after inclusion of all remaining measures of body mass and fat distribution in the model, android fat remained the only independent predictor of higher PWV and AIx and lower SEVR. Gynoid fat mass and waist circumference were negatively associated with AIx which is likely a gender effect. BMI remained independently associated with lower FMD (Table 2).

Table 2:

Multivariate Analyses at Baseline

Table 2: n = 711	RHI		PWV (m/s)		AIx		SEVR		FMD (%)
	β	p-value	β	p-value	β	p-value	β	p-value	β	p-value
Multivariate analysis 1
BMI (kg/m2)	−0.056	−0.246	0.112	0.007	−0.122	0.001	−0.158	<0.001	−0.258	<0.001
Multivariate analysis 2
BMI (kg/m2)	−0.096	0.355	−0.089	0.312	−0.168	0.029	0.204	0.029	−0.341	<0.001
Android Fat Mass (kg)	0.045	0.666	0.230	0.010	0.052	0.497	−0.414	<0.001	0.094	0.173
Multivariate analysis 3 *
BMI (kg/m2)	−0.002	0.988	0.039	0.725	0.059	0.538	0.166	0.163	−0.364	<0.001
Android Fat Mass (kg)	0.079	0.522	0.305	0.004	0.240	0.008	−0.407	<0.001	0.126	0.127

Multivariate analyses controlling traditional cardiovascular risk factors and measures of body mass. Each multivariate analysis included the following variables in addition to the listed variables in the first column: age, gender, ethnicity, tobacco use, diabetes, mean arterial pressure, current antihypertensive or statin use, total cholesterol, HDL-c, and C-reactive protein. Standardized beta coefficient displayed with p-value.

^*Multivariate analysis 3 controlled for waist circumference, waist:hip ratio, and gynoid fat mass in addition to aforementioned variables.

Table 4:

One-year follow-up controlled for baseline and change in traditional risk factors and change in gynoid fat

Table 4: n = 498	ΔMAP (mmHg)		ΔSEVR		ΔFMD (%)
	β	p-value	β	p-value	β	p-value
ΔBMI (kg/m2)	−0.013	0.776	−0.166	<0.001	−0.248	<0.001
ΔAndroid:Gynoid Ratio	−0.011	0.823	−0.131	0.006	−0.124	0.009
ΔAndroid Fat (kg)	0.045	0.339	−0.124	0.009	−0.124	0.009

Multivariate analyses of one-year change were controlled for gender, ethnicity, tobacco use at baseline, baseline age, baseline hypertension, baseline diabetes, change in tobacco use at one year, change in HDL, change in total cholesterol, change in gynoid fat mass, and new statin, antihypertensive, or diabetic medication use after one year.

At one year, 498 of the subjects returned for repeat testing with similar baseline characteristics. More subjects were diagnosed with or treated for hyperlipidemia (n=16 on statins), diabetes mellitus (n=4) or hypertension (n=11) at one year (Table 1). After one year, subjects lost a mean 1.3 kilograms in weight that corresponded with significant reductions in BMI, waist circumference, waist:hip ratio, and android and gynoid fat mass. Subjects on average had reduced systolic and diastolic blood pressures and improved total cholesterol and LDL at one year (Table 1, Supplementary Table 1). After one year, there were significant reductions in arterial stiffness (PWV, AIx), but also significant reductions in FMD and RHI (Table 1).

At one-year follow-up, the change in BMI, gynoid, and android fat mass were positively associated with change in mean arterial pressure. Moreover, the change in BMI, the android:gynoid ratio, and android fat correlated with change in SEVR (r= −0.25, p <0.001; r= −0.19, p <0.001; r= −0.24, p <0.001; respectively) and FMD (r= −0.16, p <0.001; r= −0.14, p= 0.003; r= −0.12, p= 0.013; respectively) (Table 3). Change in measures of fat distribution were not associated with corresponding changes in RHI, PWV, and AIx (Supplementary Table 4). Thus, reductions in fat were associated with reduced arterial stiffness, improved endothelium-dependent vasodilation and reduced blood pressure.

Table 3:

Bivariate Analyses of one-year follow-up

Table 3: n = 498	ΔMAP (mmHg)		ΔSEVR		ΔFMD (%)
	r	p-value	r	p-value	r	p-value
ΔBMI (kg/m2)	0.124	0.006	−0.246	<0.001	−0.162	<0.001
ΔGynoid Fat (kg)	0.132	0.004	−0.200	<0.001	−0.060	0.201
ΔAndroid:Gynoid Ratio	0.052	0.263	−0.187	<0.001	−0.140	0.003
ΔAndroid Fat (kg)	0.141	0.004	−0.238	<0.001	−0.116	0.014

Bivariate correlations between changes in fat distribution and measures of arterial function and stiffness at one year. One-year change in mean arterial pressure is also included as measures of arterial stiffness correlate with blood pressure. Measures of change are the difference between the valve at one-year and at baseline.

After adjusting for changes in gynoid fat mass as well as changes in traditional cardiovascular risk factors and changes in diabetic, antihypertensive, and antihyperlipidemic medications after one year, change in BMI, the android:gynoid ratio, and android fat remained negatively correlated with changes in SEVR (β= −0.16, p <0.001; β= −0.13, p= 0.006; β= −0.12, p= 0.009; respectively) and FMD (β= −0.25, p <0.001; β= −0.12, p= 0.009; β= −0.12, p= 0.009; respectively) (Table 4). Android fat and BMI remained the only significant predictors of SEVR and FMD in this multivariate model. Although change in fat mass positively correlated with change in blood pressure (Table 3), increases in BMI and android fat distribution were more closely related to increases in arterial stiffness and reduction in endothelium-dependent vasodilation.

As the measures of fat distribution were highly correlated, we performed additional analyses of lower risk and non-obese subgroups. An analysis of non-obese subjects, those with baseline BMI <30 kg/m², demonstrated similar results. The same analysis was performed on females alone, subjects with 10-year ASCVD risk ≤5%, and subjects <60 years-old with BMI <28 kg/m² revealing similar results. Additionally, analyzing fat distribution as a ratio of BMI or by a percent of total weight (e.g. android:BMI ratio, android percentage of total weight) produced similar results.

Discussion:

In one of the largest studies analyzing the relationships between fat distribution measured by DXA and several measures of vascular function, we found that android, but not gynoid fat distribution is an important and independent determinant of arterial stiffness, even after adjustment for all other covariates. These findings were strengthened by observations after one-year follow-up, where an increase in android fat mass but not gynoid fat was associated with an increase in arterial stiffness and impairment of endothelium-dependent vasodilation, and vice versa. BMI and android fat mass were highly correlated. Although BMI remains an independent predictor of arterial stiffness and endothelial function, its association with vascular dysfunction appears to be importantly driven by the magnitude of android fat.

Android fat or abdominal obesity is increasingly recognized as a contributor to cardiovascular disease. Abdominal obesity is an independent predictor of cardiovascular risk and has recently been described to have an additive impact on cardiovascular risk in patients with hypertension.²⁷ Visceral adiposity has been associated with concentric left ventricular remodeling and adverse hemodynamics.²⁸ Furthermore, abdominal obesity has adverse relationships with subclinical non-invasive markers of arterial health (e.g. PWV,^29–32 AIx,³³ and SEVR³⁴). Direct measurements of android and gynoid fat mass by DXA, as in our study, have shown similar results as an elevated android:gynoid fat ratio is associated with cardiometabolic dysregulation³⁸ and increased android fat mass is associated with increased fasting blood glucose, blood pressure, total cholesterol, triglycerides, and reduced physical activity.³⁹ Our study uses DXA allowing direct measurement of tissue density and confirms the negative associations between android fat mass and arterial health in a large population throughout the lifespan with relatively few co-morbidities.

Although there is much literature on cross-sectional analyses of fat distribution and cardiovascular risk, few studies have analyzed the effects of changes in fat distribution over time. Visceral fat influences cardiovascular risk through paracrine and endocrine mechanisms related to the metabolic syndrome increasing circulating inflammatory cytokines contributing to systemic inflammation.³⁵ Whether abdominal obesity can be a modifiable risk factor of cardiovascular disease has been less well described. From the Framingham Heart Study, increasing visceral fat over 5–6 years was associated with a higher likelihood of hypertension, hyperlipidemia, and metabolic syndrome.¹² Few registries have directly measured android and gynoid fat distributions and analyzed the effects of differing distributions of fat over time. Android and gynoid fat distributions are known to change with age as postmenopausal women and elderly men collect larger deposits of android fat which is thought to be related to hormonal changes.^{36, 37} Interestingly, in our study, an increase in android fat but not gynoid fat was associated with increased arterial stiffness and impaired endothelial function highlighting the dynamics of arterial health. The pro-inflammatory and oxidized milieu associated with metabolic syndrome may be related to an individual’s total android fat mass. Over several years, higher android fat mass may lead to hypertension, hyperlipidemia, and metabolic syndrome confirming previous findings by computed tomography.

Overall, in our cohort, subjects lost a mean 1.3 kg in one year. Interestingly, overall PWV, AIx, and SEVR improved though FMD and RHI did not (Table 1). On the other hand, FMD was not associated in the cross-sectional multivariate models though was related to changes in fat distribution at one year (Tables 2–4). Our observational substudy of the larger Emory Predictive Health study is subject to many confounding and competing effects as there are many reasons subjects may have lost weight during the study period. The overall improvement in measures of arterial stiffness (PWV, AIx, and SEVR) may be related to weight loss while lack of improvement in measures of microvascular and endothelial function (FMD and RHI) may be more dependent on worsening risk factors (age, increased incidence of diabetes, etc.) during the study period. Importantly, analyzing the change in these measures with respect to the changes in fat distribution allows us to assess these changes with respect to one another prospectively. In these analyses, increases in BMI and android fat mass corresponded with increased arterial stiffness (SEVR) and worsening endothelium-dependent vasodilation (FMD) even after controlling for changes in gynoid fat (Table 4). We hypothesize that adipokines affect endothelial-mediated nitric oxide release and that other measures of arterial stiffness (i.e. PWV, AIx) will be affected with longer follow-up.

Body mass, as defined by BMI, is an important initial determinant of arterial health. As prior evidence suggests, the distribution of fat, particularly visceral or android fat, corresponds with increased cardiovascular risk. Fat distribution may determine how body mass influences arterial health. In our study, DXA differentiates fat mass from lean mass in pre-specified distributions where the android region is defined by the iliac crest.¹⁷ In addition to BMI, direct measurement of android fat by DXA may assist in cardiovascular risk assessment. However, more importantly, our study suggests that measurement and surveillance of abdominal obesity may have additional utility for cardiovascular risk stratification even in a healthy population. In this regard, available measures of abdominal obesity should be prospectively evaluated for their utility in cardiovascular risk modification.

Our study has several limitations. Nearly two-thirds of the subjects in cohort were women and we did not have complete follow-up. Moreover, android fat is highly associated with BMI which may result in collinearity in multivariate models. Additionally, one-year follow-up may be too short a time frame to fully analyze fat distribution’s effect on vascular health. As each subject was assigned a health partner to assist in healthy living strategies, volunteers returning for follow-up are subject to confounding. The health partner did improve the cardiometabolic risk profile of the participants in the study as evidenced by improved lipid profile, blood pressure, BMI and a decrease in tobacco use at follow-up as previously described.⁴⁰ However, these findings were across all subjects returning for follow-up. Larger cohorts with longer follow-up directed towards measuring changes in fat distribution are warranted.

In a large group of middle aged subjects with few co-morbidities, we show that an increase in BMI and android fat are associated with worsening arterial stiffness (SEVR) and reduced FMD at one year. Further studies are needed to determine the impact of android fat mass and the android:gynoid fat ratio on cardiovascular risk modification.

Highlights – Changes in Body Fat Distribution Predict Arterial Health.

• We serially measured fat distribution and arterial health in a healthy cohort.

• After one year, increased android fat correlated with worsened arterial function.

• Rather than body weight alone, fat distribution influences arterial health.

• Direct measurement of fat distribution may assist cardiovascular risk assessment.