Adipose tissue depots as predictors of coronary events in the general population

Background The link between obesity and cardiovascular disease (CVD) is complex involving not only excess body weight but also the regional distribution and quality of adipose tissue. Knowledge gap remains regarding the association between regional body fat and incident coronary events in a large general population sample. Purpose To assess whether computed tomography (CT)-derived measures of regional adipose tissue amount and quality improve the prediction of incident coronary events in the general population. Specifically, we evaluated their predictive value: (i) compared with traditional anthropometric obesity measures, (ii) compared with steatotic liver disease, and (iii) compared with, or in combination to, the systematic coronary risk evaluation 2 (SCORE2) algorithm. Methods 29385 participants from the Swedish Cardiopulmonary Bioimage Study (SCAPIS) (Reference #1) were included in the analyses. The amount (cm2) and attenuation (Hounsfield Units, HU) of subcutaneous (SAT, thigh and abdominal), visceral (VAT) and intermuscular (thigh) fat depots as well as liver attenuation were assessed from CT slices. Body mass index (BMI) and waist circumference were quantified and SCORE2 was calculated based on conventional risk factors. Steatotic liver disease was defined as liver attenuation of ≤48 HU. Coronary events data were extracted from national registers. Cox regression was used to identify predictors of incident coronary events. Predictive performance was assessed using extreme gradient boosting derived ROC curves and AUC values. Results During a median follow-up of 7.8 years, 391 coronary events occurred. Extreme gradient boosting identified VAT area, VAT to SAT abdominal area ratio, and SAT thigh HU as the three fat-related variables with the strongest predictive performance of coronary events. Model D which included these top three fat variables together with age and sex, showed better predictive performance of coronary events compared to Model B (BMI, waist circumference, age, and sex) with an AUC of 0.79 versus 0.76 (p<0.001) (Figure 1). Model D also outperformed Model C (steatotic liver disease, age, and sex) with an AUC of 0.79 versus 0.72 (p<0.001). When the top three fat variables were added to age, sex, and SCORE2 (Model E), predictive performance improved compared with Model A (SCORE2 alone), with an AUC of 0.83 versus 0.78 (p<0.001). In sex-stratified analyses, Model D showed better predictive performance than Model A in women (AUC 0.86 versus 0.83, p<0.001), but not in men (Figure 2). Conclusions In a general population without previous CVD, three CT-derived adipose tissue variables of VAT and SAT, improved discrimination of coronary events compared to anthropometric obesity metrics and steatotic liver disease, and added predictive value to SCORE2. The top three fat variables improved predictive performance of coronary events compared to SCORE2 in women, but not in men.