Epicardial and pericardial fat are distinguished as two functionally distinct cardiac depots, with deep-learning segmentation and standardization advancing their imaging quantification toward cardiovascular risk stratification

The review clearly distinguishes EAT from PAT: EAT is intrapericardial visceral fat in direct contact with the myocardium, originating from the splanchnopleuric mesoderm with brown-fat-like properties; PAT lies outside the parietal pericardium, originates from white adipose tissue, is metabolically closer to subcutaneous fat, and is supplied by the pericardiophrenic artery rather than the coronary circulation. Addressing inconsistent terminology such as 'pericardial fat' in the literature, the review adopts a definition consistent with Bertaso et al., using the pericardium as the anatomical boundary (1.0-4.0 mm thick on imaging) and noting there is no adipose tissue between the pericardial layers. An anatomical and embryological synthesis within a narrative review, based on prior literature and a schematic (Figure 1), without new primary measurements.

Physiologically and pathologically, EAT acts on the myocardium and coronary arteries through paracrine and vasocrine effects, secreting pro-inflammatory cytokines such as IL-6, TNF-alpha, and MCP-1 as well as protective adipokines such as adiponectin, with its function shifting from protective to pro-inflammatory under conditions such as obesity, insulin resistance, or type 2 diabetes; PAT acts more through systemic metabolic and inflammatory signaling. It explicitly separates EAT and PAT into 'local' versus 'systemic' mechanistic pathways and links them respectively to coronary artery disease, atrial fibrillation, and heart failure (particularly heart failure with preserved ejection fraction). An integrative discussion of prior mechanistic literature citing specific molecules (IL-6, TNF-alpha, MCP-1, adiponectin, TGF-beta, leptin) and processes such as atrial fibrosis and myocardial stiffness, without new experimental data.

The review summarizes conflicting evidence on EAT and cardiovascular outcomes: Tanami et al. found no significant association between EAT volume and severity of coronary calcification, stenosis, or myocardial perfusion abnormalities, and Mahabadi et al. observed an association only in patients under 55 with low baseline calcification; in contrast, studies by Chu, Tscharre, Lu, Eisenberg, and West showed that higher EAT thickness or volume significantly correlates with cardiovascular events. Through Table 1, it lists the largest EAT studies of the past decade by CT, echocardiography, and MRI, placing contradictory conclusions side by side and noting that EAT 'quality' metrics (CT attenuation, MRI T1 time, fat attenuation index, EAT dispersion) may carry more prognostic value than volume alone. Evidence comes from multiple meta-analyses and observational studies, with Table 1 listing large pooled samples such as 43,113, 41,534, and 352,275 participants, most associations with P values below .05; the review itself notes that heterogeneous imaging and measurement protocols limit comparability.

On treatment, the review summarizes drug and lifestyle effects on EAT: the standardized mean difference was -0.195 (95% CI -0.79, -0.32, P<.001) for statins, -1.005 (95% CI -1.37, -0.64, P<.001) for GLP-1 receptor agonists, and -0.552 (95% CI -0.79, -0.32, P<.001) for SGLT2 inhibitors, with GLP-1 receptor agonists about twice as effective as SGLT2 inhibitors (P=.04); the Mediterranean diet outperformed a low-fat diet in reducing EAT. It compares effect sizes across drug classes and lifestyle interventions side by side and introduces the SUMMIT trial cardiac MRI substudy, in which tirzepatide significantly reduced left ventricular mass and total paracardiac fat volume but showed no specific effect on EAT, suggesting EAT may be a secondary marker rather than an independent therapeutic target. Drug effect sizes come from meta-analyses of 3 studies (603 patients), 7 studies (240 patients), and 8 studies (221 patients) respectively, with limited sample sizes; the SUMMIT substudy points in a different direction, and the review presents this tension candidly.

This article is aimed at clinicians, radiologists, and medical imaging algorithm researchers interested in cardiovascular risk stratification. It applies to settings where one must choose an EAT/PAT quantification modality, understand each modality's thresholds and limitations, or assess the conditions for deploying deep-learning segmentation in clinical practice; it also suits readers wanting to understand how statins, GLP-1 receptor agonists, SGLT2 inhibitors, and lifestyle interventions affect EAT volume. The review makes clear that standardized scanning and reconstruction protocols, along with a clear distinction between EAT and PAT, are prerequisites for translating these findings into clinical practice.

Readers should still note: whether EAT and PAT are independent cardiovascular risk factors or modifiable therapeutic targets remains, by the review's own account, incompletely established with conflicting findings across studies; the reproducibility of EAT 'quality' metrics (CT attenuation, FAI, MRI T1, EAT dispersion) across platforms and populations is still under investigation; the SUMMIT substudy suggests tirzepatide did not specifically affect EAT, pointing in a different direction from some drug meta-analyses, and its mechanisms and clinical implications require further study; and although deep-learning segmentation reports high Dice coefficients, scarce training data, model interpretability, and overfitting risk remain open issues.