Epicardial Fat Thickness and Bone Mineral Content: The Healthy Twin Study in Korea

Background The conventional concept of positive association between general obesity and bone health was challenged in recent studies reporting the different effects of specific fat deposition on bone health. In the present study, we investigated the association between epicardial fat and bone health. Methods We measured echocardiographic epicardial fat thickness (EFT) and bone mineral content (BMC) in a twin-family cohort of Koreans (n = 1,198; 525 men, 460 pre- and 213 post-menopausal women). A total 121 pairs of monozygotic twin (MZ) and 404 pairs of dizygotic twin and sibling pairs (DZ/Sib) were included. Results EFT was positively associated with BMC in total, as well as in three subgroups (β = 0.107, 0.076, and 0.058 for men, pre-, and post-menopausal women, respectively). The positive association between EFT and BMC remained for DZ/Sib difference analysis, but was absent for MZ comparisons. The positive association between BMI and BMC was consistent for DZ/Sib and MZ difference analysis. After adjusting for the effect of general obesity via BMI, the association between BMC and EFT was statistically non-significant (β = 0.020, 0.000, and −0.009 for men, pre-, and post-menopausal women, respectively). Conclusion Our findings do not support epicardial fat’s beneficial effects on bone health, whereas general adiposity has an osteotropic effect. The association between EFT and BMC is through common genetic component factors.


INTRODUCTION
Obesity and osteoporosis are significant public health problems with increasing prevalence and substantial economic burdens in most industrialized countries. [1][2][3][4] Although many studies have reported a relationship between bone health and obesity, the association is still inconclusive. Previous epidemiologic studies have demonstrated that higher body mass index (BMI) and body weight had protective roles against bone loss, and weight reduction was associated with significant bone loss. [5][6][7] In this setting, obesity and fat seems to have beneficial effect on bone health in contrast to the detrimental effect on most health conditions, especially on cardiovascular and metabolic diseases. 8,9 However, since anthropometric data, such as BMI and body weight, reflect lean mass as well as fat mass (FM), the actual effect of the fat tissue on bone heath is not clear. Therefore, the role of fat on bone health needs to be investigated with a new indicator.
Several recent studies have been shedding new light on the complex bone-fat connection. Adipocytes and osteoblasts seem to interact and reciprocally modulate at many levels, including the human mesenchymal stem cells in the bone marrow, from which they both originate. 10,11 In addition, a recent publication demonstrated that different regional fat depots have different impacts on bone health. For example, visceral adipose tissue (VAT) might have detrimental effect on bone mass, in contrast to the beneficial effect of subcutaneous adipose tissue (SAT). 12,13 Therefore, the association of bone health with more specific VAT might help to elucidate the observed osteogenic nature of obesity.
Epicardial adipose tissue (EAT) is located between the myocardium and the visceral pericardium, and has the same embryologic origin as intraabdominal mesenteric and omental fat cells. 14 Its measurement using transthoracic echocardiography (TTE) is simple and is known to be a strong predictor of abdominal VAT. 15,16 However, compared with other VAT, EAT contain more smaller adipocytes and secrete more bioactive metabolites, such as adipokines. 17,18 Recent study reported an inverse relationship between EAT and bone mineral density (BMD) in acromegaly patients, no studies have evaluated this association in the general population. 19 In the present study, we aimed to investigate the association between bone health and EAT, as well as other well-known VAT markers, including anthropometric data and regional FM measured using dual-energy X-ray absorptiometry (DXA).

Study design and population
The subjects included in this analysis were participants in the Healthy Twin Study, which was a nationwide population-based cohort study implemented as a part of the Korean genome epidemiology study. It was initiated in 2005, and participants continue to receive follow-up examinations every 3 years. Participants consisted of a twin pair and their first-degree family members. All participants received medical examinations and completed detailed questionnaires about lifestyle and epidemiologic information at one of three medical school-affiliated hospitals. Details on the study design and protocol were previously published. 20 Among the initial 1,467 subjects who completed an echocardiogram and body composition measurements between 2006 and 2008, 269 subjects were excluded: 220 for poor echocardiographic image quality, such as poor echo window or angle difference, and 49 for a treatment history of osteoporosis. A total of 1,198 subjects (525 men, 460 premenopausal women, and 213 postmenopausal women) were included in our final analysis with monozygotic twin pairs (MZ) (n = 121 pairs) and pooled dizygotic twin and sibling pairs (DZ=Sib) (n = 404 pairs). Women were considered postmenopausal if they had no history of menstruation during the previous year and fulfilled at least one of the following criteria: natural menopause, use of estrogen replacement therapy, or age older than 50 years. 21,22 Natural menopausal women was defined as those with at least 12 consecutive months of amenorrhea not due to surgery and other obvious cause, such as medical treatment or breastfeeding, which is mainly caused by loss of ovarian function. In fact, surgical menopausal women may have normal ovarian function. Therefore, in present study, women with surgical amenorrhea were considered to be postmenopausal women only if they fulfilled the history of bilateral oophorectomy or estrogen replacement therapy or they were 50 years of age or older, which can explain the loss of ovarian function.
All participants provided written informed consent. The study protocol was approved by the Institutional Review Board at Seoul National University School of Public Health.

Measurement of epicardial fat thickness
Subjects underwent TTE according to standard techniques in the left lateral decubitus position, using commercially available instruments (Vivid E9; GE Healthcare, Horten, Norway). The images were recorded in a digital database. The measurement of epicardial fat thickness was performed by one internist and one cardiologist using an offline DICOM (Digital imaging and Communications in Medicine) viewer (Onis 2.5 professional version; Digital Core, Tokyo, Japan). They were unaware of the subjects' clinical information.
Epicardial fat was identified as the echo-free space between the myocardium and the visceral epicardium, and its thickness was measured perpendicularly on the free wall of the right ventricle at end-systole from the standard parasternal long axis view. 23,24 In order to standardize the measurements between the observers, the aortic annulus was used as an anatomical landmark and the epicardial fat thickness was measured at the point on the free wall of the right ventricle along the midline of the ultrasound beam, perpendicular to the aortic annulus ( Figure 1). The intra-and inter-observer correlations for the measurement of epicardial fat thickness were good, with intraclass correlation coefficients of

Measurement of anthropometric data and body composition
Body weight and height were measured according to standard methods while the subjects were wearing a light gown or light indoor clothing. Minimum waist circumference (WC) was measured in the standing position at the point between the lower rib margin and the iliac crest. Hip circumference was measured as the largest circumference over the buttock. BMI (kg=m 2 ) was calculated as weight divided by height squared, and the waistto-hip ratio (WHR) was calculated as WC divided by hip circumference. Whole body bone mineral content (BMC; kg), BMD (g=cm 2 ) of the whole body, the lumbar spine, and the pelvis; whole-body mass, regional FM (kg), and lean mass (LM; kg) were measured using DXA (Delphi W; Hologic, Boston, MA, USA).

Clinical information
The following clinical and demographic data were extracted from each patient's baseline questionnaire: past medical history of chronic diseases, including hypertension, diabetes mellitus, hyperthyroid disease, and osteoporosis; female reproductive history, including age at menopause and use of estrogen replacement therapy; and information about cigarette smoking, alcohol consumption, and exercise habits.

Statistical analysis
The value of each continuous variable was expressed as a mean (standard deviation [SD]). Each categorical or discrete variable was presented as a percentage. Comparisons among the groups (men, premenopausal women, and postmenopausal women) were performed using analysis of variance, analysis of covariance (ANCOVA), and the Chi-squared (χ 2 ) test. Multiple comparisons between two groups were performed using post hoc analysis. The relationships between the epicardial fat thickness (EFT) and other measures of body composition were analyzed using Pearson's or Spearman's correlation analysis. Linear mixed models in each group were used to evaluate associations between EAT and bone mass, and to correct for familiar interdependence. Age and height were included as covariates in regression model 1, and past medical history (hypertension, diabetes mellitus, and thyroid disease) as well as behavioral factors (smoking, alcohol, and exercise habits) were added as covariates in model 2. In order to evaluate the exact relationship between EAT and bone mass, general obesity markers, such as BMI (model 3) or total FM (model 4), were added as covariates for excluding general obesity effect of fat on bone. Age, height, BMI, past medical history variables, and lifestyle variables were adjusted as fixed effects, and each family and twin unit was adjusted as a random effect in our linear mixed models. Additionally, we conducted within-pair analysis for MZs, same-sex DZs, and age adjusted same-sex sibling pairs. By comparing general population, MZ and DZ=sibling, we could gain further insight into the nature of the association, particularly that of genetic correlation or environmental correlation. 25,26 Because MZ shares 100% of genetic information, any meaningful differences within MZ pairs are reasonably interpreted as nongenetic contributions including epigenetics. On the other hand, if associations in general populations are materially weakened or nullified in the MZ comparisons, it strongly suggests that the associations might stem from common genetic grounds. These findings are further supported by the positive associations in the DZ pair comparisons, who share only 50% of genetic constitutions but similar level of environmental sharing. In the present study, differences in obesity indices within pairs were compared with the differences in the BMC within the same pairs using Spearman's correlation coefficient.
Statistical analyses were performed using Statistical Package for the Social Sciences statistical software (version 18; SPSS-IBM, Chicago, IL, USA) or by R version 3.02 (R Foundation for Statistical Computing, Vienna, Austria). All tests were two-tailed, and a P-value <0.05 was considered statistically significant.

RESULTS
The baseline characteristics of the subjects are listed in Table 1. The body composition parameters and clinical information were statistically different according to sex and menopausal status. Men had higher BMI, WC, WHR, and LM than women. Fatrelated parameters, such as total FM and trunk FM, were lower in men than in women, with the exception of head FM. Postmenopausal women had more total fat and trunk fat, but had lower leg fat and LM than premenopausal women. EFT was highest in postmenopausal women and lowest in premenopausal women. BMC was higher in men-because of their larger body size-than in women. However, the difference in BMD between the sexes was small, and no significant difference in spine BMD was identified between men and premenopausal women. Hypertension and diabetes mellitus were most prevalent in postmenopausal women; risky heath behaviors, such as smoking and drinking alcohol, were the highest in men. Table 2 and eFigure 2 show the correlations of EFT with anthropometric and body composition variables. EFT was highly We also examined the differences in whole-body BMC, wholebody BMD, and body-part specific BMD across the tertiles of EFT, adjusting for age and height in all three subgroups using the ANCOVA test. As shown in Table 3, BMC significantly increased across increasing tertiles of EFT, especially in men and postmenopausal women. This tendency was not observed in whole-body or spine BMD, particularly in women. The results of the post hoc analysis between tertiles are shown in eTable 1. Figure 2 shows the association between BMC and EFT and trunk fat. In bivariate unadjusted analyses, BMC increased with increasing EFT in all subgroups. These tendencies were also observed in associations between BMC and trunk fat in all subgroups. Table 4 shows the multivariate-adjusted associations between BMC and body composition variables after correcting for familial interdependence. EFT was positively associated with BMC in men, premenopausal women, and postmenopausal women in the age-and height-adjusted model (model 1). This association was consistent after controlling for past medical history (hypertension, diabetes, and hyperthyroid disease) and behavioral factors (smoking, alcohol, and exercise habits; model 2). Trunk FM and classical indices of abdominal obesity parameters, including WHR, were also positively associated with BMC in both models. The same associations were observed with total FM. Model 3 shows the associations between BMC and EFT and other body composition variables after adjusting for general obesity with BMI. The association between BMC and EFT was statistically insignificant in all groups. In model 4, which adjusted for general obesity using total FM, there was a significant positive correlation between epicardial fat and BMC in men, but no statistical significant association was observed in premenopausal women or postmenopausal women. However, the association between BMC and the conventional abdominal obesity markers, including WHR and trunk FM converted to negative after additionally adjusting for BMI. Table 5 shows the within-pair analysis in MZ and DZ=Sib. The positive association between EFT and BMC was consistent in DZ= Sib difference analysis, while within-pair difference analysis for MZ, which shared 100% of their genetic constitutions, cancelled   The unit of beta coefficient for BMC and epicardial fat thickness is gram= mm; Waist-to-hip ratio and BMC, gram=0.1 unit; trunk fat mass and BMC, gram=kilogram; soft lean mass, gram=kilogram.
Epicardial Fat and Bone this association. This strongly suggested an association of genetic nature between the BMC and EFT. Contrary to EFT, the positive association between BMI and BMC was consistent for both DZ= Sib and MZ in difference analysis, indicating either genetic or environmental associations. The results according to sex and menopausal status in within-pair analysis are shown in eTable 2. Approximately 15% of the initial subjects were excluded from analysis due to poor echocardiographic images that affected measurement reliability (eFigure 3). These subjects were younger and had lower total fat (including epicardial fat thickness) and abdominal fat than the subjects with good echocardiographic images (eTable 3). The associations between BMC and EFT did not change when subjects with poor echocardiographic images were included in the analysis. The β coefficient of the association between EFT and BMC in Model 2 was 0.117 for men (P = 0.000), 0.076 for premenopausal women (P = 0.000), and 0.065 for postmenopausal women (P = 0.003).  DZ=Sib difference: pooled same-sex dizygotic twins and age-adjusted samesex sibling pairs, where the pairwise differences in obesity measures were regressed on the differences in the BMC of the same pairs.

DISCUSSION
The results from our study of the Korean Healthy Twin cohort did show that EFT was positively associated with BMC, regardless of sex and menopausal status. However, after additionally adjusting for general obesity, the association was statistically insignificant for most outcomes. Additionally, although abdominal obesity parameters, such as WHR and trunk FM, were positively associated with BMC, this association shifted to a negative or insignificant correlation after adjusting for general obesity with BMI or total FM. In addition, we demonstrated that the association between the BMC and EFT had a significant genetic basis.
The exact association between fat and bone density is still controversial. Several previous studies reported positive associations between body fat and BMC or BMD, and two plausible mechanisms were suggested on the basis of two main characteristics of fat. 27 One mechanism is related to increased weight bearing of bones, which directly activates adaptive bone remodeling [28][29][30] ; the other mechanism is associated with paracrine and hormonal effects of fat, which enhance anabolic effects on bone through increased production of sex hormones and hormonal factors, such as insulin, leptin, and amylin. [31][32][33][34][35] However, other previous reports demonstrated a negative relationship between fat and BMC or BMD. 7,[36][37][38] In most of these studies, body weight was used as an important covariate in the analysis; however, this may have created a false association between FM and bone mass due to biases from strong co-linearity between FM and body weight. 39 Therefore, we used height as an important covariate instead of weight; because whole-body BMC and BMD are highly associated with whole-body bone size and height is known to be a good surrogate marker for body size. 27,36 Also, the negative association between fat and bone might be partially due to the specific role of different region's fat on bone. For example, several studies showed that abdominal visceral fat was negatively associated with bone mass, in contrast to the positive association of subcutaneous fat with bone. 12,13 Epicardial fat, located only between the myocardium and the visceral pericardium, is known to have considerable systemic effects by secreting bioactive adipokines and being involved in the lipid metabolism, such as in the production of free fatty acid. 40 Until now, studies on epicardial fat have been focused on cardiovascular diseases and metabolic syndrome, but there have been few studies on the relation between epicardial fat and bone health, although adipokines secreted from the fat are significantly associated with bone metabolism. 41,42 A recent study showed an inverse relationship between epicardial fat and BMD of lumbar spine in acromegaly patients, which was mediated by Dickkopfrelated protein 1 (DKK1), an inhibitor of osteoblast differentiation and bone formation produced by preadiopocytes and osteocytes in humans. 19 However, no studies have evaluated this association in the general population.
In the present study, EFT was positively associated with BMC. The association of other conventional abdominal fat variables, including WHR and trunk FM, with BMC was consistent with this result. However, after adjusting for general obesity with BMI, the positive association between EFT and BMC converted to negative, though it was statistically insignificant. In addition, the associations between BMC and conventional abdominal obesity variables also converted to negative. Kim et al reported that abdominal visceral fat had an important role in negatively regulating bone mass if the weight-bearing effect of fat was excluded, which is consistent with our results. 43 However, given that the directionality between bone, general obesity (or FM), and EFT are still unclear, we cannot exclude the possibility that BMI works as a collider rather than a pure confounder, which our model assumed. But, it is unlikely that bone health status significantly affects BMI or other obesity indices, because fat is metabolically and biologically much more active than the bones and its influences are exerted on virtually all body systems. Bones are also a metabolically active tissue, but the regulations and influences are overwhelmingly toward bone and bone mineral homeostasis. Given this biology, we believe the adjustment of BMI might be insufficient or too simple for testing independent EFT effects, but it is unlikely that BMI works as a collider between EFT and BMC. In addition, considering that the results are similar when adjusting for general obesity using total FM, which does not act as a collider, BMI is also unlikely to act as a collider in this study.
To date, no study addressing the genetic correlation between EFT and BMC has been conducted. Analyzing twins and sibling pairs in the Korean Healthy Twin study cohort, the positive correlation between EFT and BMC was absent in within-pair difference analysis for MZ, which suggests the involvement of genetic constitutions in these findings. It is imperative to conduct further studies investigating the common genetic determinants. In addition, this association suggests that genetically identical individual's bone health might not benefit from the increase in EFT level.
This study has several strengths. Principally, we used multiple methods for estimating different types of adiposity and covariates that influence BMC. In addition, having a family-and-twin structure enabled us to dissect the associations between fat measures and BMC into those of genetic and non-genetic nature.
However, this study has several limitations. This study had a cross-sectional design, and all participants were Korean. Additionally, echocardiographic EFT may not reflect the exact quantity of total epicardial fat because it is a linear measurement and varies at different locations around the myocardium.
In conclusion, this study suggests that general obesity contributes to the observed positive association between EFT and BMC. The lack of correlation between BMC and EFT independent of BMI or total FM does not support the possible beneficial role of EFT in bone health. However, given the complex inter-correlation between general obesity EFT and bone health, findings from simple adjustment might not preclude an association between EFT and BMC. Moreover, our findings suggest that the observed association between EFT and BMC might involve a genetic correlation.