Abstract
Background:
Visceral fat has been shown to be associated with increased cardiometabolic risk, but the role of subcutaneous fat remains unclear, and evidence from diverse populations is lacking. We hypothesized that visceral fat, but not subcutaneous fat, would be independently associated with incident cardiovascular disease (CVD) and all-cause mortality.
Methods:
Among 1910 participants from the Multi-Ethnic Study of Atherosclerosis with abdominal fat measurements from computed tomography scans and followed for an average of 9.3 years, we used multivariable Cox proportional hazards models to investigate the relationship of both visceral and subcutaneous fat tertiles with CVD and all-cause mortality. We tested for interaction and performed sensitivity analysis for subgroups and missing values of visceral fat.
Results:
Participants had mean age of 65 years, visceral fat 150 cm2, subcutaneous fat 263 cm2, and 50% were female, 21% African American, 13% Asian, and 26% Hispanic. In models adjusted for age, sex, race/ethnicity, income, education, smoking, and subcutaneous fat, there was a statistically significant positive association between visceral fat and CVD, but not mortality. The association for combined CVD may be driven by incident coronary heart disease [tertile 2: hazard ratio, 2.43 (1.38 to 4.28); tertile 3: hazard ratio, 3.00 (1.66 to 5.43)]. Additionally, we found no substantial associations between subcutaneous fat and CVD or mortality. There were no statistically significant interactions by age, sex, or race/ethnicity.
Conclusions:
Visceral fat, but not subcutaneous fat, is significantly associated with increased risk for CVD in a multiethnic cohort. These data support the need for effective strategies for lifestyle changes that prevent and reduce visceral fat.
Using survival analysis, we found that visceral fat, but not subcutaneous fat, was associated with incident cardiovascular disease; these estimates were similar across age, sex, and race/ethnicity.
There is accumulating evidence that high levels of visceral fat are associated with incident cardiovascular disease (CVD) and mortality, and that the relationship for visceral fat is stronger than that for subcutaneous fat (1–8). Very few studies have investigated whether this relationship is specific to the type of CVD event, and risk-based clinical cut-points for visceral fat have yet to be determined.
It has been suggested that subcutaneous fat is not associated with, or even may be protective for, CVD risk (9–12). One theory of ectopic fat accumulation supporting this hypothesis is that fat is deposited in the subcutaneous compartment first and then in the visceral cavity only when subcutaneous storage capacity is exceeded (13, 14). Although the ability to preferentially place fat in the subcutaneous cavity may be beneficial, higher subcutaneous fat may still be associated with increased cardiometabolic risk, especially if ectopic fat deposition becomes more likely as the subcutaneous capacity is being approached.
The evidence for a relationship between visceral and subcutaneous fat with incident CVD or mortality comes predominantly from small studies in white populations (1, 2, 5–7). We are aware of only two prior studies, from the same cohort, that have investigated whether these relationships differ between individuals self-reporting white and African American race/ethnicity (3, 15). Similarly, there are two studies in participants of Japanese ancestry (4, 8) and, to our knowledge, none from a Hispanic cohort. This lack of evidence from diverse populations limits the generalizability of prior findings.
Given the evidence from the existing literature, we hypothesized that higher visceral fat would be significantly associated with higher risk for incident CVD events and all-cause mortality, but that any associations with subcutaneous fat would be explained by the correlation with visceral fat. To add evidence to the field, we investigated these associations in specific CVD events, including coronary heart disease (CHD), stroke, and heart failure, and also determined if these relationships differed by age, sex, and race/ethnicity.
Methods
Study population
From 2000 through 2002, the Multi-Ethnic Study of Atherosclerosis (MESA) enrolled 6814 participants ages 45 to 84 years free of CVD at six sites across the United States (16). Participants were racially/ethnically diverse (white, 39%; Asian, 12%; African American, 28%; Hispanic, 22%). Clinic visits were conducted approximately every two years. For a 30% random subsample from all MESA sites (n = 1947), computed tomography (CT) scans of the abdomen were obtained and used to measure body composition during visits 2 (n = 769) and 3 (n = 1178) starting in 2002 and 2004, respectively (17). For this study, we used measurements from the L4/L5 intervertebral disc space and excluded participants who had CVD events before body composition measurement (n = 37).
Demographics of the Body Composition Ancillary Study population are similar to the overall MESA cohort. Using semiautomated Medical Image Processing, Analysis, and Visualization software, average visceral and subcutaneous fat masses were estimated from each of two slices (18). Participants with conditions precluding a scan were excluded and 24 scans were unreadable. Certain participants had one or more body cavities that did not fit in the field of view, producing missing data. If visceral fat values were missing and the visceral area was cut off on only one side (n = 25), the area was imputed by determining the measurement from the midline for the complete half of the image and then doubled. Very few visceral fat values were completely missing (n = 24); however, it is likely that the data are not missing at random (17), because the largest participants are the most likely to not fit in the field of view and also most likely to have high levels of visceral fat. Subcutaneous scans missing area only on one lateral side (n = 216) were imputed using the same half-process as for visceral fat. If more than one side of the subcutaneous area was missing, then missing values were multiply imputed using sex and race/ethnicity specific regression equations (n = 146). Values imputed from regression equations were only used for sensitivity analyses. Only 173 subcutaneous values were missing and could not be imputed using these methods. A random subset of Body Composition Ancillary Study participants (n = 590) underwent a second CT scan at visit 4 starting in 2005. Visceral and subcutaneous fat were measured in the same manner as for the first CT scan. Change over time was calculated as fat mass at the second scan – fat mass at the first scan. Visceral fat change was characterized as an increase if change was >30 cm2, a decrease if change was <−30 cm2, and no change if between −30 and 30 cm2. Subcutaneous fat change was similarly characterized.
All MESA participants were followed through 2013 for all-cause mortality and incident CVD events including CHD, stroke, heart failure, and combined CVD. Using a standardized adjudication protocol, CHD, and combined CVD events were categorized as “hard” or “all” events (listed in Supplemental Text (471.1KB, pdf) 1). Primary analysis for this study focused on all events.
Age, sex, race/ethnicity, education, income, and smoking status were self-reported at baseline; CVD risk factors such as hypertension, diabetes, total and high-density lipoprotein (HDL) cholesterol, and triglycerides were matched to the same visit as the abdominal CT scan (study visit 2 or 3). Hypertension was defined as systolic blood pressure ≥140mm Hg, diastolic blood pressure ≥90mm Hg, or use of blood pressure–lowering medications. Type 2 diabetes was defined as fasting glucose ≥126 mg/dL, or use of glucose-lowering medications. All covariates were measured using a standard protocol (16).
Statistical analysis
We described the baseline characteristics of the study sample by tertiles of visceral fat. We used Cox proportional hazards models to quantify the associations between tertiles of visceral fat and incident CVD events and all-cause mortality, with the lowest tertile as the reference group. We constructed survival times from the visit of visceral fat measurement (visit 2 or 3) to the event. We created nested Cox models to adjust for confounding: model 1 was unadjusted; model 2 adjusted for age; model 3 further adjusted for age, sex, race/ethnicity, education, income, and smoking; and model 4 included mutual adjustment for visceral and subcutaneous fat. A priori, we assumed that hypertension, type 2 diabetes, and hyperlipidemia were mediators of this relationship and therefore did not adjust for them in our primary analysis. We formally tested for effect modification by age, sex, and race/ethnicity using interaction terms. Our investigation of subcutaneous fat used the same analytic plan as that for visceral fat. Finally, we used logistic regression to assess whether change in visceral or subcutaneous fat from the first CT scan to the second CT scan was associated with CVD and mortality.
We also assessed the continuous associations of visceral fat with CVD and mortality incidence rates using adjusted Poisson regression models. For the Poisson models, we used linear splines with two knots, one at each tertile cut-point for visceral fat. To further evaluate the continuous relationships, we used the Contal and O’Quigley approach, implemented by the Cha, Mandaker, and Mandaker SAS macro, to investigate outcome-based empirical cut-points for visceral fat (19–21). All analysis, except the estimation of cut-points, was conducted using Stata 11 (22).
Sensitivity analyses
We conducted sensitivity analysis investigating the difference in the association between body composition and hard CHD and CVD events compared with all CHD and CVD events. We analyzed the impact of missing visceral and subcutaneous fat data on the results using indicator variables as markers of missingness for visceral and subcutaneous fat, both in the Cox proportional hazards models and Kaplan-Meier curves (n = 1590 for the primary model vs n = 1908 for the missing indicator model). We also investigated the a priori assumption that other CVD risk factors were mediators of the relationship between adiposity and CVD by adding a mediation model that included hypertension, diabetes, and hyperlipidemia. To determine whether visceral fat predicts CVD and mortality above and beyond the traditional measure of overall adiposity, we included body mass index (BMI) in the model and used likelihood ratio tests to assess whether BMI had an independent contribution to CVD risk. We assessed whether confounding by physical activity, measured as total intentional exercise, explained our results. Finally, we determined whether results differed for subgroups with type 2 diabetes or who used hormone replacement therapy.
Results
A total of 1910 participants were included in the analysis and followed for an average of 9.3 years. Compared with participants in the lowest tertile of visceral fat, participants in the highest tertile are older; more likely to be male; have lower income and education; have higher BMI, blood pressure, triglycerides, HbA1c, and subcutaneous fat; and lower HDL cholesterol (Table 1).
Table 1.
Baseline Characteristics of 1910 Adults Age 45–84 Years in the MESA Body Composition Ancillary Study by Visceral Fat Category
| Visceral Fat |
|||||
|---|---|---|---|---|---|
| Characteristic | Tertile 1 (<110.9) | Tertile 2 (110.9–171.9) | Tertile 3 (≥171.9) | Missing Visceral Fat | P Value for Trenda |
| N | 623 | 623 | 640 | 24 | — |
| Visceral fat, cm2 | 78.5 (0.89) | 138.8 (0.70) | 230.2 (1.86) | NA | — |
| Age, y | 63.6 (0.39) | 64.0 (0.39) | 66.0 (0.37) | 66.8 (1.87) | <0.001 |
| Female | 399 (64) | 317 (51) | 238 (37) | 4 (17) | <0.001 |
| Race | 0.34 | ||||
| White | 223 (36) | 229 (37) | 303 (47) | 14 (58) | |
| Asian | 121 (19) | 97 (16) | 30 (4.7) | 1 (4.2) | |
| African American | 184 (30) | 129 (21) | 88 (14) | 2 (8.3) | |
| Hispanic | 95 (15) | 168 (27) | 220 (34) | 7 (29) | |
| Education (% ≥ high school) | 528 (85) | 518 (83) | 500 (78) | 18 (75) | 0.001 |
| Income (% ≥ $35,000) | 359 (58) | 344 (55) | 330 (52) | 15 (63) | 0.055 |
| Current smoking, % | 80 (13) | 71 (11) | 92 (14) | 2 (8.3) | 0.54 |
| BMI, kg/m2 | 24.6 (0.16) | 28.1 (0.17) | 31.4 (0.19) | 34.6 (1.14) | <0.001 |
| Total cholesterol, mg/dL | 191.9 (1.39) | 192.4 (1.36) | 187.0 (1.48) | 176.8 (5.85) | 0.002 |
| LDL cholesterol, mg/dL | 112.4 (1.24) | 115.1 (1.21) | 110.2 (1.32) | 104.4 (4.80) | 0.083 |
| HDL cholesterol, mg/dL | 59.6 (0.66) | 50.6 (0.56) | 45.3 (0.46) | 42.1 (2.15) | <0.001 |
| Triglycerides, mg/dL | 98.8 (2.23) | 138.0 (3.54) | 162.6 (4.76) | 151.5 (11.81) | <0.001 |
| Hypertension, % | 208 (33) | 301 (49) | 362 (57) | 16 (67) | <0.001 |
| Systolic BP, mm Hg | 119.2 (0.83) | 124.6 (0.79) | 127.8 (0.83) | 133.9 (3.83) | <0.001 |
| Type 2 diabetes, % | 44 (7.1) | 83 (12) | 128 (20) | 7 (29) | <0.001 |
| HbA1c, % | 5.56 (0.05) | 5.84 (0.07) | 5.93 (0.08) | 5.96 (0.20) | <0.001 |
| Subcutaneous fat, cm2 | 220 (4.05) | 273 (4.76) | 309 (6.03) | b | <0.001 |
| Across follow-up | |||||
| CHD, % | 20 (3.2) | 45 (7.2) | 67 (10) | 5 (21) | <0.001 |
| Stroke, % | 17 (2.7) | 24 (3.9) | 21 (3.3) | 3 (13) | 0.24 |
| Heart failure, % | 18 (2.9) | 20 (3.2) | 32 (5.0) | 2 (8.3) | 0.027 |
| Combined CVD, % | 39 (6.3) | 70 (11) | 88 (14) | 8 (33) | <0.001 |
| Mortality, % | 59 (9.5) | 64 (10) | 92 (14) | 5 (21) | 0.003 |
Values are mean (standard deviation) or frequency (%).
Abbreviations: BP, blood pressure; LDL, low-density lipoprotein.
Cuzick nonparametric test for trend.
All subcutaneous fat values are missing for those missing visceral fat.
Kaplan-Meier curves for combined CVD showed that visceral fat tertiles 2 and 3 were significantly different than tertile 1 (P < 0.005), but overlapping curves for tertiles 2 and 3 (P = 0.18), suggest some evidence for a potential plateau in risk (Fig. 1A). The informative nature of the missing data at the upper range of the visceral fat distribution, with statistically significant separation of the curve for the missing values (P < 0.005), indicates that the plateau may be an artifact of the missing data. In contrast, Kaplan-Meier curves for combined CVD by tertiles of subcutaneous fat showed no separation between the curves for the tertiles or for missing values (Fig. 1B).
Figure 1.
(A) Kaplan-Meier curves for combined cardiovascular disease survival by visceral fat tertiles and missing status in 1910 MESA participants. Visceral fat tertiles: tertile 1 (<110.9 cm2), tertile 2 (110.9 to <171.9 cm2), and tertile 3 (≥171.9 cm2). Curves for tertiles 2 and 3 are not significantly different from each other but are statistically different from tertile 1 and missing values at the P < 0.05 level. Tertile 1 and missing values are also significantly different from each other. Number at risk for years 2, 4, 6, 8, and 10 is 1831, 1739, 1677, 1605, and 584. (B) Kaplan-Meier curves for combined cardiovascular disease survival by subcutaneous fat tertiles and missing status in 1910 MESA participants. Subcutaneous fat tertiles: tertile 1 (<203.7 cm2), tertile 2 (203.7 to <286.8 cm2), and tertile 3 (≥286.8 cm2). There is no statistically significant difference between any of the curves (P > 0.05). Number at risk for years 2, 4, 6, 8, and 10 is 1831, 1739, 1677, 1605, and 584.
Results from the Cox proportional hazards models showed substantial positive associations for visceral fat tertiles for CHD and combined CVD, but not for all-cause mortality (Table 2 and Fig. 2). Although there was weak evidence for a positive association with heart failure, there was no evidence of an association for stroke (Table 2). Estimates for CHD and CVD were mildly attenuated with additional adjustment, although estimates were generally stronger when adjusted for subcutaneous fat. No evidence was found for interaction between visceral or subcutaneous fat. Despite some evidence of higher estimates for women, those younger than 70 years and those self-reporting African American or Hispanic race/ethnicity for CHD and combined CVD (Table 3), there were no statistically significant interactions with age, sex, or race/ethnicity (all P > 0.15). Cox proportional hazard models also estimated no consistent statistically significant association between subcutaneous fat and any CVD event or death (abbreviated results in Supplemental Table 1 (471.1KB, pdf) ). Visceral fat was a stronger risk factor than subcutaneous fat for CHD and combined CVD (Fig. 2). There were no significant deviations from the proportionality assumption for any of the models.
Table 2.
HRs and 95% CIs for Cause-Specific CVD and All-Cause Mortality by Visceral Fat Tertiles in MESA
|
Visceral Fat Tertiles |
||||||
|---|---|---|---|---|---|---|
|
1 (<110.9 cm2) |
2 (110.9–171.9 cm2) |
3 (≥171.9 cm2) |
||||
| Model | Events | HR | HR | 95% CI | HR | 95% CI |
| CHD | ||||||
| Model 1 | 132 | 1.0 | 2.31 | 1.36–3.91 | 3.45 | 2.09–5.68 |
| Model 2 | 132 | 1.0 | 2.30 | 1.36–3.89 | 3.10 | 1.88–5.11 |
| Model 3 | 132 | 1.0 | 2.12 | 1.24–3.61 | 2.42 | 1.44–4.07 |
| Model 4 | 108 | 1.0 | 2.43 | 1.38–4.28 | 3.00 | 1.66–5.43 |
| Stroke | ||||||
| Model 1 | 62 | 1.0 | 1.45 | 0.78–2.70 | 1.26 | 0.66–2.29 |
| Model 2 | 62 | 1.0 | 1.44 | 0.78–2.69 | 1.13 | 0.60–2.15 |
| Model 3 | 61 | 1.0 | 1.33 | 0.70–2.52 | 0.98 | 0.50–1.94 |
| Model 4 | 47 | 1.0 | 1.25 | 0.61–2.54 | 0.77 | 0.33–1.77 |
| Heart failure | ||||||
| Model 1 | 64 | 1.0 | 1.28 | 0.63–2.52 | 2.00 | 1.07–3.71 |
| Model 2 | 64 | 1.0 | 1.27 | 0.64–2.49 | 1.64 | 0.88–3.06 |
| Model 3 | 64 | 1.0 | 1.12 | 0.56–2.23 | 1.16 | 0.61–2.24 |
| Model 4 | 51 | 1.0 | 1.29 | 0.61–2.75 | 1.31 | 0.60–2.84 |
| Combined CVD | ||||||
| Model 1 | 197 | 1.0 | 1.87 | 1.27–2.77 | 2.35 | 1.61–3.43 |
| Model 2 | 197 | 1.0 | 1.87 | 1.26–2.76 | 2.09 | 1.43–3.05 |
| Model 3 | 196 | 1.0 | 1.71 | 1.15–2.55 | 1.61 | 1.09–2.40 |
| Model 4 | 161 | 1.0 | 1.76 | 1.15–2.70 | 1.73 | 1.10–2.73 |
| All-cause mortality | ||||||
| Model 1 | 214 | 1.0 | 1.09 | 0.77–1.56 | 1.53 | 1.11–2.13 |
| Model 2 | 214 | 1.0 | 1.08 | 0.76–1.54 | 1.26 | 0.90–1.75 |
| Model 3 | 213 | 1.0 | 1.02 | 0.71–1.46 | 1.09 | 0.77–1.54 |
| Model 4 | 179 | 1.0 | 1.14 | 0.77–1.67 | 1.30 | 0.87–1.95 |
Model 1, unadjusted; model 2, age; model 3, model 2 + sex, race/ethnicity, education, income, and smoking; model 4, model 3 + subcutaneous fat. Bold values indicate significant difference from the reference group at the P < 0.05 level.
Abbreviations: CI, confidence interval; HR, hazard ratio.
Figure 2.
Cox proportional hazards ratios and 95% confidence intervals for incident coronary heart disease and combined cardiovascular disease by mutually adjusted visceral and subcutaneous fat tertiles in 1590 MESA participants. Adjusted for age, sex, race/ethnicity, education, income, smoking, and other fat depot. Visceral fat tertiles: 1 (<110.9 cm2), 2 (110.9 to <171.9 cm2), and 3 (≥171.9 cm2). Subcutaneous fat tertiles: 1 (<203.7), 2 (203.7 to <286.8), and 3 (≥286.8 cm2).
Table 3.
Sensitivity Analyses (HR and 95% CIs) for CHD, Combined CVD, and All-Cause Mortality by Visceral Fat Level in MESA
| CHD |
CVD |
All-Cause Mortality |
|||||
|---|---|---|---|---|---|---|---|
| Model | Visceral Fat Tertilea | Event | CI | Event | CI | Event | CI |
| HR | HR | HR | |||||
| Primary modelb | 132 | 196 | 213 | ||||
| 1 | 1.0 | 1.0 | 1.0 | ||||
| 2 | 2.12 | 1.24–3.61 | 1.71 | 1.15–2.55 | 1.02 | 0.71–1.46 | |
| 3 | 2.42 | 1.61–4.07 | 1.61 | 1.09–2.40 | 1.09 | 0.77–1.54 | |
| Hard events onlyc | 89 | 145 | NA | ||||
| 1 | 1.0 | 1.0 | |||||
| 2 | 1.92 | 1.02–3.61 | 1.59 | 1.01–2.51 | |||
| 3 | 2.41 | 1.31–4.45 | 1.64 | 1.04–2.57 | |||
| Missing datad | |||||||
| Indicators for missing visceral and subcutaneous fat | 137 | 204 | 218 | ||||
| 1 | 1.0 | 1.0 | 1.0 | ||||
| 2 | 2.19 | 1.27–3.77 | 1.67 | 1.11–2.52 | 1.06 | 0.73–1.54 | |
| 3 | 2.66 | 1.52–4.67 | 1.65 | 1.07–2.55 | 1.15 | 0.78–1.69 | |
| Missing | 5.37 | 1.62–17.78 | 4.92 | 1.87–12.96 | 1.38 | 0.47–3.95 | |
| Mediation analysise | 130 | 194 | 209 | ||||
| 1 | 1.0 | 1.0 | 1.0 | ||||
| 2 | 1.72 | 0.98–3.00 | 1.44 | 0.95–2.18 | 1.00 | 0.69–1.46 | |
| 3 | 1.69 | 0.96–2.98 | 1.19 | 0.78–1.83 | 1.01 | 0.69–1.49 | |
| Continuous associations for assessment of heterogeneity per 50 cm2 visceral fat | |||||||
| Primary modelb | 132 | 196 | 213 | ||||
| 1.15 | 1.03–1.29 | 1.08 | 0.97–1.19 | 1.02 | 0.93–1.12 | ||
| Age | |||||||
| <70 y | 71 | 103 | 72 | ||||
| 1.21 | 1.03–1.42 | 1.16 | 1.01–1.33 | 0.99 | 0.83–1.17 | ||
| ≥70 y | 61 | 93 | 141 | ||||
| 1.13 | 0.95–1.34 | 1.04 | 0.90–1.20 | 1.07 | 0.96–1.21 | ||
| P value for interaction | 0.15 | 0.27 | 0.66 | ||||
| Sex | |||||||
| Female | 44 | 73 | 79 | ||||
| 1.27 | 1.01–1.60 | 1.20 | 1.01–1.44 | 1.07 | 0.89–1.28 | ||
| Male | 88 | 123 | 134 | ||||
| 1.11 | 0.97–1.27 | 1.02 | 0.91–1.15 | 1.00 | 0.89–1.13 | ||
| P value for interaction | 0.38 | 0.24 | 0.91 | ||||
| Race/ethnicity | |||||||
| White | 66 | 93 | 98 | ||||
| 1.09 | 0.93–1.27 | 1.05 | 0.92–1.20 | 0.98 | 0.85–1.12 | ||
| Asian | 12 | 16 | 22 | ||||
| 0.95 | 0.54–1.66 | 0.81 | 0.49–1.34 | 1.09 | 0.71–1.67 | ||
| African American | 26 | 40 | 56 | ||||
| 1.23 | 0.94–1.62 | 1.04 | 0.81–1.32 | 0.92 | 0.75–1.15 | ||
| Hispanic | 28 | 47 | 37 | ||||
| 1.32 | 1.03–1.70 | 1.23 | 1.00–1.51 | 1.31 | 1.04–1.66 | ||
| P value for interaction | 0.48 | 0.83 | 0.38 | ||||
| Type 2 diabetes (diagnosis or medication use) | |||||||
| No | 108 | 162 | 176 | ||||
| 1.15 | 1.01–1.30 | 1.07 | 0.96–1.20 | 1.00 | 0.90–1.12 | ||
| Yes | 24 | 34 | 37 | ||||
| 1.05 | 0.77–1.43 | 0.99 | 0.76–1.28 | 0.95 | 0.74–1.22 | ||
| P value for interaction | 0.33 | 0.16 | 0.038 | ||||
| Current smoking | |||||||
| No | 111 | 164 | 182 | ||||
| 1.16 | 1.03–1.32 | 1.12 | 1.00–1.24 | 1.06 | 0.96–1.18 | ||
| Yes | 21 | 32 | 31 | ||||
| 0.98 | 0.72–1.35 | 0.84 | 0.65–1.10 | 0.73 | 0.54–0.98 | ||
| P value for interaction | 0.76 | 0.18 | 0.076 | ||||
| Hormone replacement (women only) | |||||||
| No | 25 | 41 | 44 | ||||
| 1.16 | 0.86–1.57 | 1.10 | 0.87–1.39 | 1.06 | 0.84–1.33 | ||
| Yes | 19 | 31 | 35 | ||||
| 1.46 | 1.03–2.09 | 1.36 | 1.04–1.79 | 1.07 | 0.81–1.43 | ||
| P value for interaction | 0.57 | 0.32 | 0.32 | ||||
Bold values indicate significant difference from the reference group at the P < 0.05 level.
Abbreviation: CI, confidence interval.
Visceral fat tertiles (1: <110.9; 2: 110.9 to <171.9; 3: ≥171.9 cm2) and subcutaneous fat tertiles (1:<203.7; 2: 203.7 to <286.8; 3: ≥286.3 cm2).
Primary model = model 3 from Table 2: age, sex, race/ethnicity, education, income, and smoking.
Hard CHD events include: myocardial infarction, resuscitated cardiac arrest, and CHD death (all CHD also includes definite or probable angina). Hard CVD includes: myocardial infarction, resuscitated cardiac arrest, CHD death, stroke, and stroke death (all CVD also includes definite or probable angina if followed by revascularization, other atherosclerotic death, and other CVD deaths).
Primary model includes nonmissing data and data where subcutaneous values could be imputed using the half-process method for a total of n = 1590. The missing data model minimizes missing data by including all possible values for missing data: subcutaneous fat imputed with prediction equations, indicator for fully missing subcutaneous fat data, and an indicator for fully missing visceral fat data. This model places completely missing values for visceral and subcutaneous fat into separate categories for a total sample size of n = 1908. Visceral fat is completely missing for n = 24; subcutaneous fat is completely missing for n = 173.
Mediation: Primary model + hypertension, type 2 diabetes, total cholesterol, HDL cholesterol, and triglycerides.
Average visceral fat change was 4.19 cm2 and average subcutaneous fat change was 3.24 cm2. Logistic regression models offer some indication that increase in total fat mass may be associated with CHD and CVD in addition to higher baseline visceral fat (Table 4). No significant association was found for visceral or subcutaneous fat change and all-cause mortality.
Table 4.
ORs and 95% CIs for CVD and All-Cause Mortality by Body Composition at Baseline and Change Over Time in MESA
| CHD |
CVD |
All-Cause Mortality |
|||||
|---|---|---|---|---|---|---|---|
| Model | OR | CI | OR | CI | OR | CI | |
| Visceral fat | |||||||
| Tertile 1 | 1.0 | 1.0 | 1.0 | ||||
| Tertile 2 | 5.08 | 1.48–17.4 | 2.61 | 1.13–6.04 | 3.15 | 1.04–9.52 | |
| Tertile 3 | 5.62 | 1.47–21.5 | 2.16 | 0.82–5.69 | 2.73 | 0.76–9.77 | |
| Visceral fat change | |||||||
| Decrease | 0.98 | 0.21–4.54 | 0.69 | 0.20–2.46 | 0.83 | 0.22–3.15 | |
| No change | 1.0 | 1.0 | 1.0 | ||||
| Increase | 1.37 | 0.51–3.70 | 0.89 | 0.39–2.00 | 0.41 | 0.09–1.81 | |
| Subcutaneous fat | |||||||
| Tertile 1 | 1.0 | 1.0 | 1.0 | ||||
| Tertile 2 | 1.18 | 0.43–3.21 | 1.04 | 0.45–2.38 | 1.51 | 0.51–4.46 | |
| Tertile 3 | 0.50 | 0.13–1.88 | 1.26 | 0.48–3.29 | 1.55 | 0.43–5.56 | |
| Subcutaneous fat change | |||||||
| Decrease | 0.88 | 0.20–3.99 | 1.29 | 0.46–3.61 | 1.80 | 0.61–5.37 | |
| No change | 1.0 | 1.0 | 1.0 | ||||
| Increase | 2.86 | 1.09–7.49 | 2.36 | 1.07–5.17 | 0.12 | 0.13–1.06 | |
All models include adjustment for age, sex, race/ethnicity, education, income, and smoking. Bold values indicate significant difference from the reference group at the P < 0.05 level. Visceral fat tertiles (1: <110.9; 2: 110.9 to <171.9; 3: ≥171.9 cm2), subcutaneous fat tertiles (1:<203.7; 2: 203.7 to <286.8; 3: ≥286.3 cm2), visceral fat change (decrease: <–30; no change: –30 to 30; increase: >30 cm2), and subcutaneous fat change (decrease: <–30; no change: –30 to 30; increase:>30 cm2).
Abbreviation: CI, confidence interval.
The adjusted continuous relationships of visceral fat with CHD and CVD incidence rate were generally linear, while the association with all-cause mortality exhibited a weak potential U-shape (Supplemental Fig. 1 (471.1KB, pdf) ). The curves for CHD and CVD did not display the plateau suggested by Fig. 1 and Table 2, further indicating that data at the upper range of visceral fat may be missing differentially. Results for the cut-point analysis are described in Supplemental Text 2 (471.1KB, pdf) .
Sensitivity analyses confirmed that results were similar for most subgroups (Table 3), with some loss of precision and monotonicity, predominantly in smaller subsamples. Adjusting for total intentional exercise did not change the results (data not shown). Similarly, including BMI in the model resulted in minor attenuation (not shown), mostly for CVD, but likelihood ratio tests for the inclusion of BMI (P = 0.17, 0.10, and 0.71) and BMI estimates (P = 0.16, 0.098, and 0.71) were not significant for CHD, CVD, or mortality, respectively. Limiting the analysis to only hard endpoints also resulted in attenuation, more pronounced for CVD than for CHD (Table 3). Including a category for missing visceral fat values indicated that the exclusion of missing values may bias the results for CHD and combined CVD. Finally, informal mediation analysis showed that accounting for mediation by the traditional CVD risk factors attenuated the results, particularly for combined CVD [hazard ratio for tertile 2 = 1.44 (0.95–2.18) and tertile 3 = 1.19 (0.78–1.83)].
Discussion
Our primary finding is that visceral fat was significantly associated with an increased risk for both CHD and combined CVD, but not all-cause mortality, whereas subcutaneous fat was not associated with any of these outcomes. These results were generally robust to sensitivity analyses; however, the missing data at the top of the visceral fat distribution limited our investigation of the true magnitudes and shapes of the observed relationships. These results did not differ significantly by age, sex, or race/ethnicity. Risk factors on the pathway partially (for CHD) or substantially (for CVD) mediated the relationship between visceral fat and cardiovascular events. There was no evidence for meaningful cut-points for visceral fat based on cardiovascular risk.
Our results are consistent with prior findings that visceral, but not subcutaneous fat, is significantly associated with incident CVD and that this association remains when BMI is included in the model (1–4). Although reports on the association between visceral fat and CVD risk factors are common, there are still only a handful of prospective longitudinal studies that have investigated the role that body fat composition plays in CVD events (1–4) or mortality (1, 5–8). Our study investigated specific CVD endpoints as well as combined CVD. As such, we provide evidence suggesting that risk estimates for combined CVD from visceral fat burden may be driven by CHD. Similarly, only one prior study specifically investigated the role that visceral fat plays in the development of heart failure and none has investigated the association with stroke (15). Larger studies or collaborations that combine cohorts may be needed to fully investigate these lower event rate outcomes.
Our finding that the association between higher visceral fat and incident CHD and CVD is similar across racial/ethnic groups is consistent with Nicklas et al., in which no substantial heterogeneity between whites and African Americans was demonstrated (3, 15), but our work adds comparable estimates in Asian and Hispanic individuals from the same cohort. Known disparities in CVD risk and reported differences in body fat distributions by race/ethnicity would make these findings unexpected (23, 24) if it were not for the similar results for visceral fat and CHD in white, African American, and Japanese-American cohorts (3, 4). The lack of biological mechanisms for such heterogeneity in conjunction with the evidence that anthropometric measures have differential ability to approximate visceral fat by race/ethnicity also make this finding less surprising (17, 25). Although the relative strength of the association of visceral fat and subcutaneous fat with CVD risk for each racial/ethnic group needs further assessment, this study bolsters support for interventions to reduce visceral fat for primary prevention in all racial/ethnic groups.
Evidence for interaction in the association between visceral fat and CVD by sex is contradictory (1, 3). Although the results of our study do not indicate heterogeneity by sex, they do support the findings by Nicklas et al. that the visceral fat and CVD relationships may persist in the elderly (Table 3). In this regard, it has been suggested that the lower estimates of the BMI vs CVD relationship in this group is due to the inaccuracy of BMI as a measurement of adiposity with advancing age. Use of the more precise visceral fat measurement that directly captures fat in the depot known to be most strongly associated with risk may ameliorate this potential misclassification. Similarly, the use of visceral fat to assess the adiposity vs CVD relationship may show that sex differences seen in these relationships when anthropometric proxies are used are due to measurement and statistical artifact; however, more studies powered to detect interaction are needed to clarify this issue.
Except for the weak and generally nonsignificant positive estimates for stroke and the protective estimates for mortality, there was little evidence in MESA that subcutaneous fat plays a role in cardiovascular risk. Consistent failure in the literature to find a protective effect from subcutaneous fat and the preferential decrease in visceral fat from weight loss (26) indicate that weight loss for those with indications of high visceral fat burden may offer the most health benefits. These results support the idea that primary prevention of risk factors and CVD may be possible through generalized weight loss without negative implications from loss of subcutaneous fat. Methods to target abdominal adiposity may provide additional tools for reduction of CHD risk.
Also in concordance with prior evidence (27), our mediation analysis suggests that obesity, including visceral fat, acts through comorbidities such as hypertension, diabetes, and hypercholesterolemia to increase CVD risk. The extent of this mediation and alternative pathways between obesity and CVD risk, especially pathways specific to fat in the visceral compartment, remain to be fully elucidated. More pertinent is that these findings further support a logical causal framework for the health effects of visceral fat specifically, and the potential for obesity prevention to impact CVD incidence.
Prior evidence on the association between body fat composition and mortality is less consistent. Contrary to our weakly suggestive, but nonsignificant results, all studies reported that visceral fat was significantly associated with increased all-cause mortality, but estimates differed widely by sex and inclusion of BMI (5–7), including two studies with estimates that were fully attenuated by the inclusion of BMI and other CVD risk factors (1, 8). Mortality results for subcutaneous fat also differed, from a substantial and persistent protective association in women only (6), to null results (1, 5, 8), to an important positive univariate association with complete attenuation by mutual adjustment for other fat depots in men (7). The heterogeneity of estimates for visceral fat and mortality may be explained by differences in the distribution and relationship of noncardiovascular causes of death or by true subgroup effects. The possibility of a null relationship for mortality does not negate the relationship between visceral fat and CHD in the context of the obesity epidemic and the associated burden and cost of resulting cardiovascular comorbidities.
Missing data at the upper ranges of visceral fat pose the most challenging limitation for our analysis. Even more problematic, the results from our sensitivity analyses indicate that these data are missing not at random and likely attenuate our estimates. In contrast, these results were expected given the previously identified limitation of the CT field of view for participants above a certain size and our prior investigation of this phenomenon (17). The known mechanism for this missing data gave us the opportunity to understand the repercussions of this differential missingness and make educated assumptions about its effects on our inference. The small number of specific CVD events also limited this study by reducing the precision and possibly the stability of the estimates for stroke and heart failure. This was compounded by the reduction of follow up time in this analysis necessitated by the measurement of body fat at visits 2 and 3 instead of at baseline. Similarly, the small number of repeat CT scans across a relatively short period limits our ability to make conclusions about the importance of change in visceral and subcutaneous fat over time separately from baseline levels. This study may also be underpowered to assess quantitative interaction.
These limitations are offset by several strengths. The MESA abdominal body composition ancillary study provided the opportunity to investigate the role of adiposity in CVD risk in a large diverse cohort. Using CT-derived measurements of visceral and subcutaneous fat allowed us to separate the components of adiposity to determine the univariate and mutually adjusted associations that place people most at risk for CVD. Using these gold standard measurements also allowed us to provide more precise estimates of the effects of body fat composition on CVD risk. Similarly, the comprehensive and adjudicated reporting of CVD events and mortality allowed us to examine these relationships for specific CVD event types. Finally, the diversity of the MESA study allowed us to assess interaction by age, sex, and race/ethnicity, whereas the richness of the data allowed us to investigate the sensitivity of our results. These strengths make this study one of the most comprehensive investigations into the role that body fat composition plays in cardiometabolic risk to date.
Visceral fat, but not subcutaneous fat, was significantly associated with increased CHD and combined CVD risk in MESA, and this association was partly mediated by intermediate risk factors for CVD. There was no substantial association between visceral or subcutaneous fat and mortality. These associations did not differ by age, sex, or race/ethnicity. Reduction of obesity in general, and methods to prevent and reduce of visceral fat specifically, may lead to lower risk for CHD in diverse populations of middle-aged adults.
Acknowledgments
The authors thank the other investigators, staff, and participants of the Multi-Ethnic Study of Atherosclerosis (MESA) study for their valuable contributions. A full list of participating MESA investigators and institutions can be found at http://www.mesa-nhlbi.org. The information contained here was derived in part from data provided by the Bureau of Vital Statistics, New York City Department of Health and Mental Hygiene.
Financial Support: This research was supported by contracts HHSN268201500003I, N01-HC-95159, N01-HC-95160, N01-HC-95161, N01-HC-95162, N01-HC-95163, N01-HC-95164, N01-HC-95165, N01-HC-95166, N01-HC-95167, N01-HC-95168, N01-HC-95169, and HL088451 from the National Heart, Lung, and Blood Institute (NHLBI) and UL1-TR-000040 and UL1-TR-001079 from the National Center for Research Resources. M.M.-C. was supported by training grant NHLBI 5T32HL007261-34.
Acknowledgments
Disclosure Summary: The authors have nothing to disclose.
Footnotes
- BMI
- body mass index
- CHD
- coronary heart disease
- CT
- computed tomography
- CVD
- cardiovascular disease
- HDL
- high-density lipoprotein
- MESA
- Multi-Ethnic Study of Atherosclerosis.
References
- 1.Britton KA, Massaro JM, Murabito JM, Kreger BE, Hoffmann U, Fox CS. Body fat distribution, incident cardiovascular disease, cancer, and all-cause mortality. J Am Coll Cardiol. 2013;62(10):921–925. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Figueroa AL, Takx RA, MacNabb MH, Abdelbaky A, Lavender ZR, Kaplan RS, Truong QA, Lo J, Ghoshhajra BB, Grinspoon SK, Hoffmann U, Tawakol A. Relationship between measures of adiposity, arterial inflammation, and subsequent cardiovascular events. Circ Cardiovasc Imaging. 2016;9(4):e004043. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Nicklas BJ, Penninx BW, Cesari M, Kritchevsky SB, Newman AB, Kanaya AM, Pahor M, Jingzhong D, Harris TB; Health, Aging and Body Composition Study . Association of visceral adipose tissue with incident myocardial infarction in older men and women: the Health, Aging and Body Composition Study. Am J Epidemiol. 2004;160(8):741–749. [DOI] [PubMed] [Google Scholar]
- 4.Fujimoto WY, Bergstrom RW, Boyko EJ, Chen KW, Leonetti DL, Newell-Morris L, Shofer JB, Wahl PW. Visceral adiposity and incident coronary heart disease in Japanese-American men. The 10-year follow-up results of the Seattle Japanese-American Community Diabetes Study. Diabetes Care. 1999;22(11):1808–1812. [DOI] [PubMed] [Google Scholar]
- 5.Katzmarzyk PT, Mire E, Bouchard C. Abdominal obesity and mortality: the Pennington Center Longitudinal Study. Nutr Diabetes. 2012;2:e42. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Koster A, Murphy RA, Eiriksdottir G, Aspelund T, Sigurdsson S, Lang TF, Gudnason V, Launer LJ, Harris TB. Fat distribution and mortality: the AGES-Reykjavik Study. Obesity (Silver Spring). 2015;23(4):893–897. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Kuk JL, Katzmarzyk PT, Nichaman MZ, Church TS, Blair SN, Ross R. Visceral fat is an independent predictor of all-cause mortality in men. Obesity (Silver Spring). 2006;14(2):336–341. [DOI] [PubMed] [Google Scholar]
- 8.McNeely MJ, Shofer JB, Leonetti DL, Fujimoto WY, Boyko EJ. Associations among visceral fat, all-cause mortality, and obesity-related mortality in Japanese Americans. Diabetes Care. 2012;35(2):296–298. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Golan R, Shelef I, Rudich A, Gepner Y, Shemesh E, Chassidim Y, Harman-Boehm I, Henkin Y, Schwarzfuchs D, Ben Avraham S, Witkow S, Liberty IF, Tangi-Rosental O, Sarusi B, Stampfer MJ, Shai I. Abdominal superficial subcutaneous fat: a putative distinct protective fat subdepot in type 2 diabetes. Diabetes Care. 2012;35(3):640–647. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Narumi H, Yoshida K, Hashimoto N, Umehara I, Funabashi N, Yoshida S, Komuro I. Increased subcutaneous fat accumulation has a protective role against subclinical atherosclerosis in asymptomatic subjects undergoing general health screening. Int J Cardiol. 2009;135(2):150–155. [DOI] [PubMed] [Google Scholar]
- 11.Porter SA, Massaro JM, Hoffmann U, Vasan RS, O’Donnel CJ, Fox CS. Abdominal subcutaneous adipose tissue: a protective fat depot? Diabetes Care. 2009;32(6):1068–1075. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.McLaughlin T, Lamendola C, Liu A, Abbasi F. Preferential fat deposition in subcutaneous versus visceral depots is associated with insulin sensitivity. J Clin Endocrinol Metab. 2011;96(11):E1756–E1760. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Demerath EW, Sun SS, Rogers N, Lee M, Reed D, Choh AC, Couch W, Czerwinski SA, Chumlea WC, Siervogel RM, Towne B. Anatomical patterning of visceral adipose tissue: race, sex, and age variation. Obesity (Silver Spring). 2007;15(12):2984–2993. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Després JP. Is visceral obesity the cause of the metabolic syndrome? Ann Med. 2006;38(1):52–63. [DOI] [PubMed] [Google Scholar]
- 15.Nicklas BJ, Cesari M, Penninx BW, Kritchevsky SB, Ding J, Newman A, Kitzman DW, Kanaya AM, Pahor M, Harris TB. Abdominal obesity is an independent risk factor for chronic heart failure in older people. J Am Geriatr Soc. 2006;54(3):413–420. [DOI] [PubMed] [Google Scholar]
- 16.Bild DE, Bluemke DA, Burke GL, Detrano R, Diez Roux AV, Folsom AR, Greenland P, Jacob DR Jr, Kronmal R, Liu K, Nelson JC, O’Leary D, Saad MF, Shea S, Szklo M, Tracy RP. Multi-Ethnic Study of Atherosclerosis: objectives and design. Am J Epidemiol. 2002;156(9):871–881. [DOI] [PubMed] [Google Scholar]
- 17.Mongraw-Chaffin M, Golden SH, Allison MA, Ding J, Ouyang P, Schreiner PJ, Szklo M, Woodward M, Young JH, Anderson CAM. The sex and race specific relationship between anthropometry and body fat composition determined from computed tomography: evidence from the Multi-Ethnic Study of Atherosclerosis. PLoS One. 2015;10(10):e0139559. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Center for Information Technology, National Institutes of Health. MIPAV (Medical Image Processing, Analysis, and Visualization) software, 4.1.2. Bethesda, MD; 1999.
- 19.Contal C, O’Quigley J. An application of changepoint methods in studying the effect of age on survival in breast cancer. Comput Stat Data Anal. 1999;30:253–270. [Google Scholar]
- 20.Williams B, Jayawant NM, Sumithra JM, Cha SS, Furth AF. Finding optimal cutpoints for continuous covariates with binary and time-to-event outcomes. Rochester, MN: Mayo Clinic Department of Health Services. Available at: http://www.mayo.edu/research/documents/biostat-79pdf/DOC-10027230. Accessed 17 August 2017.
- 21.Mandrekar JN, Mandrekar SJ, Cha SS. Cutpoint Determination Methods in Survival Analysis using SAS®. SAS Users Group International Proceedings 28; March 30–April 2, 2003; Seattle, WA. Paper 261-28.
- 22.Stata Statistical Software. Release 14. College Station, TX: StataCorp LP; 2009.
- 23.Writing Group Members, Mozaffarian D, Benjamin EJ, Go AS, Arnett DK, Blaha MJ, Cushman M, Das SR, de Ferranti S, Després J-P, Fullerton HJ, Howard VJ, Huffman MD, Isasi CR, Jiménez MC, Judd SE, Kissela BM, Lichtman JH, Lisabeth LD, Liu S, Mackey RH, Magid DJ, McGuire DK, Mohler ER, Moy CS, Muntner P, Mussolino ME, Nasir K, Neumar RW, Nichol G, Palaniappan L, Pandey DK, Reeves MJ, Rodriguez CJ, Rosamond W, Sorlie PD, Stein J, Towfighi A, Turan TN, Virani SS, Woo D, Yeh RW, Turner MB, American Heart Association Statistics Committee; Stroke Statistics Subcommittee. Heart disease and stroke statistics—2016 update: a report from the American Heart Association. Circulation. 2016;133(4):447–454. [DOI] [PubMed] [Google Scholar]
- 24.Carroll JF, Chiapa AL, Rodriquez M, Phelps DR, Cardarelli KM, Vishwanatha JK, Bae S, Cardarelli R. Visceral fat, waist circumference, and BMI: impact of race/ethnicity. Obesity (Silver Spring). 2008;16(3):600–607. [DOI] [PubMed] [Google Scholar]
- 25.Lear SA, Humphries KH, Kohli S, Birmingham CL. The use of BMI and waist circumference as surrogates of body fat differs by ethnicity. Obesity (Silver Spring). 2007;15(11):2817–2824. [DOI] [PubMed] [Google Scholar]
- 26.Chaston TB, Dixon JB. Factors associated with percent change in visceral versus subcutaneous abdominal fat during weight loss: findings from a systematic review. Int J Obes. 2008;32(4):619–628. [DOI] [PubMed] [Google Scholar]
- 27.Lu Y, Hajifathalian K, Ezzati M, Woodward M, Rimm EB, Danaei G; Global Burden of Metabolic Risk Factors for Chronic Diseases Collaboration (BMI Mediated Effects) . Metabolic mediators of the effects of body-mass index, overweight, and obesity on coronary heart disease and stroke: a pooled analysis of 97 prospective cohorts with 1.8 million participants. Lancet. 2014;383(9921):970–983. [DOI] [PMC free article] [PubMed] [Google Scholar]