Among these healthy, recently postmenopausal women, even after accounting for overall and central adiposity, intra-hepatic fat was associated with greater insulin and hs-CRP levels, while TAT (EAT more so than PAT) was associated with adverse lipid levels and markers of glucose homeostasis, as well as with the presence of CAC. The finding that intra-hepatic fat was more strongly related to both CRP and insulin than cardiac fat, whereas cardiac fat was more strongly associated with CAC supports our hypothesis that intra-hepatic fat predisposes to inflammation and insulin resistance, while cardiac fat predicts atherosclerosis.
CRP is synthesized by the liver, while insulin resistance is now known to be a primary trigger of hepatic fat accumulation. As such, prior studies have shown that subjects with nonalcoholic fatty liver disease (NAFLD) have higher levels of both CRP18
and insulin resistance19
compared to those without NAFLD. Conversely, Kim et al. demonstrated similar CIMT values among metabolic syndrome-negative women with vs. without NAFLD.20
The present study extends these findings by also showing a relationship between intra-hepatic fat attenuation and both CRP and insulin resistance, while also demonstrating null findings between intra-hepatic fat and CIMT, in a population of women with low prevalence of both cardiometabolic risk markers and NAFLD (3%). These associations persisted even after accounting for possible confounders, such as race-ethnicity, physical activity level, and BMI.
Few studies have compared the associations of total TAT, EAT, and PAT. In these healthy, recently postmenopausal women, we found that total TAT, an additive variable of EAT and PAT measures, was associated both with CVD risk markers and CAC, and that this appeared to be driven slightly more so by associations with EAT than with PAT. When EAT and PAT were examined separately, associations with both CVD risk markers and CAC were more consistent for EAT. However, it should be noted that in the case of CAC, the magnitude of odds ratios were similar between EAT and PAT, but only achieved statistical significance for EAT. Among 111 consecutive asymptomatic patients undergoing CAC scanning (mean age 60 ± 10 years), EAT was more strongly associated with the presence and severity of CAC than either total TAT or PAT, even after adjustment for multiple CVD risk markers and BMI.21
Limited prior research from the Framingham Offspring and EISNER (Early Identification of Subclinical Atherosclerosis using Noninvasive Imaging Research) studies have compared a measure of total TAT (a combined measure that includes, in part, both PAT and EAT as we have defined them) to what corresponds to our measure of EAT (fat within the pericardial sac). Among older Framingham Offspring participants (mean age approximately 60 ± 9 years), total TAT was more strongly associated with CVD risk markers than was EAT, but EAT was more strongly associated with CAC and CVD events than was total thoracic fat, and many of these relationships were stronger in women, among whom there was more EAT than PAT.10, 11
Specifically, more EAT was associated with a higher odds of CAC after adjustment for traditional risk factors, abdominal visceral fat, BMI, and waist circumference, while the association with total TAT was no longer statistically significant after these adjustments.10
Similarly, although neither EAT nor total TAT was associated with incident CVD events independent of traditional CVD risk markers, after initial adjustment for age, BMI, and waist circumference, more EAT was statistically significantly associated with a higher risk for CVD events, while total thoracic fat was not.11
Similarly, among EISNER participants (age 45–54 years for men and 55–64 years for women with at least 1 CVD risk factor), EAT, but not total TAT, significantly improved the sensitivity and overall accuracy for the prediction of major adverse cardiac events when added to a predictive model that included CAC.22
Consistent with our findings, these publications suggest the possibility of a stronger local cardiac effect for EAT, specifically. However, two very recent, though small (n=113 and 49, respectively) studies suggest stronger associations between cardiometabolic risk factors and PAT than EAT.23, 24
These discordant results may result from the very high proportion of men vs. women in these most recent studies. Although sex-specific data are limited, men appear to have more PAT than EAT, while the reverse appears true for women, and as noted above in the Framingham Study, associations with EAT, at least, appear to differ between men and women.
EAT is metabolically active and releases a number of pro-atherogenic adipocytokines, including IL-6, TNF-α, and MCP-1,25
and there is evidence for diffusion of adipocytokines from epicardial adipocytes into the coronary artery.26
In addition, EAT has greater fatty acid synthesis and breakdown, and greater insulin-induced lipogenesis than do other fat depots, including PAT.8
These properties of EAT may contribute to vascular inflammation and coronary atherosclerosis.
The measures of subclinical atherosclerosis used in this study, CIMT and CAC, revealed differences in their associations with the ectopic fat indices. Null findings were present between all ectopic fat measures and CIMT, whereas significant associations between cardiac fat and CAC were found. These results are similar to those demonstrated in the Multiethnic Study of Atherosclerosis (MESA), which found that higher cardiac fat was significantly associated with CHD and CAC after adjustment for traditional CVD risk markers and BMI, but was not associated with CIMT independent of CVD risk markers.16, 27, 28
This pattern might be expected in light of the local paracrine effects noted above for EAT and the anatomical locations of CAC (coronary arteries) versus CIMT (carotid arteries) measurements. Similarly, the local nature of the diffusion of adipocytokines from epicardial adipocytes into the coronary artery would also make plausible the stronger associations we demonstrated between EAT and coronary atherosclerosis than for abdominal (waist circumference) or overall (BMI) adiposity. A direct local effect is further supported by our finding that EAT was associated with CAC independent of lipid, glucose, and insulin values.
Although intra-hepatic fat was more strongly associated with both hsCRP and insulin compared to EAT, in contrast to our hypothesis, our results show that EAT had the strongest association with glucose and triglyceride levels. The current findings also demonstrated that EAT was more strongly related to LDL-C as compared with PAT or intra-hepatic fat. These results are surprising, since intra-hepatic fat is believed to be in direct pathophysiologic relation to dysregulation of glucose and lipid metabolism. The finding that cardiac fat is associated with lipids and glucose, however, is not unprecedented. In a sample of 127 consecutive patients, Wang, et al demonstrated a correlation between EAT and both glucose and triglyceride levels in analyses adjusted for age and gender.15
Similar associations were reported after age adjustment in healthy post-menopausal women in the PROSPECT study.29
Among 71 men and women who underwent cardiac CT and coronary angiography for effort angina evaluation, EAT was found to be related to LDL-C in unadjusted analysis.14
In the present study, EAT remained associated with CAC after adjustment for lipids and glucose.
Although the current study utilized CT for quantification of cardiac fat, it is relevant to the potential clinical implications of cardiac fat that relationships have also been observed between echocardiographic cardiac fat and both cardiometabolic risk factors and coronary artery disease. A recent review paper by Iacobellis and Willens suggests specific cut-off values ranging from ≥3 mm to ≥11 mm, depending on the outcome, for the thickness of the echo-free space between the outer wall of the myocardium and the visceral layer of the pericardium observed using two-dimensional echocardiography for potential clinical utilization in the identification of patients at high risk of metabolic syndrome, insulin resistance, coronary artery disease, low coronary flow reserve, or subclinical atherosclerosis.30
Some limitations of this study must be taken into consideration. The design of this study was cross-sectional, so the results do not imply causality. Our study sample also included only healthy, recently menopausal women. Therefore, our results may not be applicable to an older population with greater prevalence of atherosclerosis. Additionally, because glucose was measured during screening, while insulin was obtained from blood collected at the baseline randomization examination, we were unable to estimate insulin resistance using the homeostasis model assessment (HOMA-IR). Instead, we have used glucose and insulin levels separately as indicators of glucose homeostasis.
One of the strengths of this study is the relatively large sample size from the KEEPS trial. In addition, this is the largest study to have simultaneously reported associations with total TAT, EAT, and PAT, and the first to compare the strength of each of these associations to those for intra-hepatic fat in the same sample of individuals. Further, because healthy women of various body sizes and education levels were enrolled in this study, the results may apply broadly to recently postmenopausal women. Lastly, our tests of reliability confirmed that each of the non-invasive CT-derived measures of ectopic fat are highly reproducible, suggesting the possibility of clinical use.
In conclusion, these analyses in healthy, midlife women suggest that while intra-hepatic fat is more closely related to measures of inflammation and insulin, cardiac fat is substantially more strongly associated with CAC. BMI and WC, measures of general adiposity, were more closely related to cardiometabolic risk factors than any ectopic fat measure, but their association with both CIMT and CAC was null. Taken together, this study suggests that cardiac fat may be a useful marker for increased CVD risk beyond the standard adiposity measures BMI and waist circumference.