Post-mortem case characteristics

The key anthropometric characteristics, heart weight metrics, and adipose volumes for the total post-mortem case cohort cases are shown in Table 1. Causes of death included cardiovascular disease/event (N = 45), subarachnoid or subdural hemorrhage (N = 4), alcohol/drug toxicity (N = 12), hanging (N = 7), asphyxiation (N = 2), asthma (N = 2), gastrointestinal hemorrhage (N = 3), sepsis (N = 2), diabetic ketoacidosis (N = 2), or other (N = 8). Further cause of death detail can be found in our previous study of this case cohort [17].

Table 1 Post-mortem case characteristics, unindexed heart weight, and unindexed adipose volumes measured from post-mortem computed tomography

Post-mortem cases had a median age of 56 (range of 18–86 years), 28% were female, and the median BMI was 28.0 kg/m2 (range, 13.0–48.9 kg/m2), indicating an overall middle-aged, male-dominant, overweight case cohort. The median heart weight was 435 g; however, there was a wide range of heart weights in the cohort (215–865 g). A similar range was found for the EAT volume (range, 12–221 cm3; median, 66 cm3) measured from PMCT and the derived EAT mass (range, 11–204 g; median, 61 g), as well as the ePAT and VAT volumes. Despite the wide variability in heart weights and EAT volumes, no outlier values were identified by ROUT analysis; all values were maintained in the dataset after outlier analysis.

Univariable associations of body size metrics and adipose volumes with heart weight

Spearman correlation analysis was used to determine the univariable associations of post-mortem case characteristics with heart weight (see Table 2 for correlation coefficients). Female sex was negatively associated with heart weight, while age had no association. As expected [20], unindexed heart weight was positively and robustly associated with anthropometric metrics, including body height and weight, BMI, BSA, and FFM. EAT volume and EAT mass (derived from EAT volume), as well as ePAT and VAT volumes were also significantly and positively associated with unindexed heart weight (Table 2); although, the strength of the ePAT association was notably less. These univariable data indicate that heart weights measured from our post-mortem cohort are associated with classic anthropometric indices and that EAT deposition is associated with heart weight.

Table 2 Univariable correlation analyses for heart weight predictorsMultivariable associations of body size metrics and adipose volumes with heart weight

Stepwise linear regression was used to determine whether classic anthropometric indices and different adipose volumes are predictors of unindexed heart weight in our case cohort independent of age and sex (Model 1). Each classic body size measurement and adipose volume, including EAT volume, was associated with unindexed heart weight independent of age and sex (Model 1, Table 3). Of the classic predictors, BSA and FFM could explain 54% and 53% of total heart weight variation, when modelled alongside age and/or female sex. EAT volume and female sex could explain 35% of heart weight variation, while ePAT and VAT volumes could only explain 17% and 23% of heart weight variation, respectively, when modelled with the female sex (Table 3).

Table 3 Stepwise linear regression for predictors of heart weight including EAT volume

Next, EAT volume, ePAT volume, or VAT volume were added to each model separately to determine whether their association with heart weight is independent of classic body size predictors (Model 2). EAT volume was associated with heart weight independent of body height, body weight, BSA, FFM, BMI, ePAT volume, or VAT volume (Table 3). Moreover, the inclusion of EAT volume improved the R2 adjusted value for each model. The association of ePAT volume with heart weight was maintained after adjusting for body height and BMI, but not for body weight, BSA, FFM, or EAT volume (Table S1). The association of VAT with heart weight was independent of body height, but not of body weight, BSA, FFM, BMI, or EAT volume (Table S1). Together, these data confirm that EAT volume is associated with heart weight independently of age, sex, classic body size predictors, and other visceral adipose volumes.

Differential associations with heart weight in cardiac hypertrophy

The simple linear regression of EAT volume as a predictor of heart weight can be appreciated in Fig. 1 (R2 for total cohort = 0.25). However, it becomes apparent that more of the variation in the EAT volume association with heart weight is found in post-mortem cases with large heart weights or EAT volumes. To explore this further, we categorized the cases into those with (median heart weight, 528 g; range, 375–865 g) or without (median, 380 g; range, 215–500 g) cardiac hypertrophy according to contemporary heart weight prediction tools [20]. Relative to the cases without hypertrophy, the cases with hypertrophy, especially those with an extreme heart weight or EAT volume, diverge considerably from the regression line (Fig. 1a). This divergence is observationally greater than that for the heart weight association with classic predictors BSA or FFM (Fig. 1b and c). Limitations in heart weight estimation from a known EAT volume are also observable in the Bland-Altman plot in Fig. 1D, whereby the estimation accuracy is diminished at the extremes of heart weight values. Again, this pattern is more pronounced for EAT volume estimation of heart weight than for estimation from BSA or FFM (Fig. 1e and f).

Fig. 1

Association and prediction of heart weight with EAT volume, BSA, and FFM. a–c simple linear regression analyses of EAT volume (a), BSA (b), and FFM (c) as univariable predictors of heart weight. d–f Bland-Altman plots showing heart weight prediction accuracy of estimates derived from simple linear regressions of the EAT volume (d), BSA (e), and FFM (f) associations with heart weight. All analyses were performed on raw (non-transformed) data. EAT, epicardial adipose tissue; BSA, body surface area; FFM, fat-free mass. N = 87

The non-hypertrophic and hypertrophic group characteristics are shown in Table S2. There was no difference in the median age or proportion of females. The hypertrophic group had significantly greater median body weight, BMI, BSA, FFM, ePAT volume, and VAT volume (Table S2). The median EAT volume (Fig. 2a) and EAT mass (Table S2) were approximately 1.9-fold greater in the hypertrophic group.

Fig. 2

EAT/myocardium mass ratio differences in cardiac hypertrophy. a Median EAT/myocardium mass ratios in the total post-mortem cases and cases without hypertrophy (non-hypertrophied, NHT, grey) and with hypertrophy (HT, red). b Scatter plot showing association between EAT/myocardium mass ratio and heart weight in NHT and HT groups. For a, the difference between NHT and HT groups was determined using the Mann-Whitney U test. NHT N = 43, HT N = 44

These univariable data indicate that hypertrophy status is an important confounding variable in the EAT volume association with heart weight. We then confirmed this using stepwise linear regression incorporating EAT volume alongside age, female sex, and different body size metrics (Table 4). In each model, EAT volume was independently associated with heart volume. However, this was only present in non-hypertrophied cases. Interestingly, models incorporating EAT volume with BSA or FFM and age or female sex could explain up to 86% of heart weight variation in non-hypertrophied cases.

Table 4 Stepwise linear regression for predictors of heart weight in post-mortem cases with and without cardiac hypertrophyEAT/myocardium mass ratio in non-hypertrophied and hypertrophied hearts

Next, we used EAT mass estimated from the EAT volume to determine if the EAT/myocardium mass ratio differs in hypertrophied hearts. In the total post-mortem case cohort, the median EAT/myocardium mass ratio was 13.4% with a range from 3.3 to 36.6% (Fig. 2b). The median EAT/myocardium mass ratio was significantly greater in hypertrophied hearts (15.5%) relative to non-hypertrophied hearts (12.5%), despite having considerably greater variation (hypertrophied, 3.6–36.6%; non-hypertrophied, 3.3–16.5%). The variation in EAT/myocardium mass ratio for a given heart weight in hypertrophic cases can also be appreciated in the scatter plot in Fig. 2c. Simple regression analysis identified hypertrophy classification as a significant univariable predictor of the EAT/myocardium mass ratio alongside age, ePAT volume, and VAT volume (Table 5). Stepwise linear regression incorporating all variables found that age, ePAT volume, VAT volume, as well as female sex were the only independent predictors (R2 = 0.45) (Table 5). These data indicate an imbalance in EAT and myocardial composition of the heart in cardiac hypertrophy.

Table 5 Regression analyses for predictors of the EAT/myocardium mass ratio in post-mortem cases