Between January 2019 and June 2022, a retrospective cohort study including 107 consecutive patients who underwent CCTA followed by invasive coronary angiography (ICA) within 3 months at The Second Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China, was performed. A total of 129 healthy controls were also included in the study to compare the CCTA-derived parameters of the patients with those in the normal population. A flowchart of this study is illustrated in Fig. 1.

All participants were of Asian ethnicity. The inclusion criteria were as follows: (1) heart failure defined according to the European Society of Cardiology (ESC) guidelines [2]; (2) New York Heart Association [NYHA] class II-IV or signs consistent with the Framingham criteria; (3) no coronary artery stenosis on coronary angiography; and. (4) CT-FFR > 0.8. The exclusion criteria included (1) missing images or poor image quality; (2) incomplete laboratory tests; and (3)) incomplete ultrasound data and loss to follow-up. All patient information, including demographic characteristics, medical history, laboratory tests, and echocardiography results, was collected after admission.

All patients provided informed consent. This study was approved by the Human Research Ethics Committee of the Second Affiliated Hospital, Zhejiang University School of Medicine (I20221200).

Medical records including medical history, laboratory data, medication, and clinical course were reviewed. Laboratory blood analysis was performed within 48 h before CCTA. Based on previous literature, metabolic syndrome was diagnosed in patients with a body-mass index (BMI) greater than 25 kg/m2 and two or three of the following conditions [15]: (1) fasting plasma glucose ≥ 110 mg/dl or HbA1c (NSGP) ≥ 5.5%, (2) systolic blood pressure ≥ 130 mmHg or diastolic blood pressure ≥ 85 mmHg, and (3) triglycerides ≥ 150 mg/dl or high-density lipoprotein (HDL) cholesterol < 40 mg/dl. Throughout the entire echocardiographic assessment, a single-lead electrocardiogram was recorded. Following the guidelines of the American Society of Echocardiography, a phased-array transducer with a fusion frequency of 2–4 MHz was employed [16]. Imaging was conducted in M-mode and 2D modes, capturing images in parasternal long-axis, short-axis, apical four-chamber, and two-chamber views and at the aortic root for evaluation. Measurements included the thicknesses of the interventricular septum (IVS), the anterior wall of the left ventricle (LVID), and the posterior wall of the left ventricle (LVPW) during both systole and diastole of the heart. The diameter of the connection point between the aorta and the sinus of Valsalva (AO-stj) and anteroposterior diameter of the left atrium (LA-ap) were also measured. LVEF was calculated using the biplane Simpson's method, averaged over three consecutive heartbeats. BMI was calculated as weight (kilograms) divided by height (meters) squared.

All participants underwent dual-source CT scans (Somatom Definition Flash or SOMATOM Force, Siemens Healthcare, Germany), which included coronary artery calcium scoring (CACS) and CCTA. The following parameters were used for the CACS protocol: tube voltage, 120 kV; tube current, 80 mAs with automated tube current modulation (CARE Dose 4D, Siemens Healthineers); and section thickness, 2 mm with a 1.5 mm increment. CCTA was conducted with a prospective ECG-gated sequence: CARE kV (reference tube voltage, 100 kV);); CARE Dose 4D (reference tube current, 288 mAs); acquisition phase, 30–75%; 65% R-R interval; and reconstructed slice thickness, 0.75 mm with a 0.5 mm increment. The images were reconstructed by advanced model iterative reconstruction (ADMIRE) level 3; kernel Bv40. Contrast medium was administered at a dose ranging from 30 to 60 mL, and the flow rate ranged from 4.0 to 6.0 mL/s according to the patients’ BMI, heart rate, and tube voltage. CCTA was performed after the injection of contrast medium (iodine 370 mg/mL [Ultravist, Bayer Schering Pharma, Berlin, Germany]) via an 18–20 G intravenous catheter, followed by a 30 mL saline flush using a dual-head power injector.

The CT-derived EAT (CTEAT), mass (CTMASS), V/M (CTV/M), FFR (CTFFR) and PCAT (CTPCAT) were assessed on a workstation (syngo.via VB40, Siemens Healthineers, Germany). Two radiologists who were blinded to the patient data (SX Lin [observer 1] and CJ Liu [observer 2], with over 5 years of experience in CT diagnostic imaging). CTFFR was calculated and analysed based on a software prototype (cFFR, syngo.via Frontier, version 3.0.1, Siemens Healthineers, Germany).

CTEAT, defined as the total amount of adipose tissue between the surface of the heart and the visceral layer of the pericardium, was measured by volumetry on short-axis slices with a thickness of 0.75 mm, ranging from the level of the pulmonary bifurcation to the apex and within a threshold range of -190 to -30 Hounsfield units (HU) to determine the total volume of tissue [17]. The adipose tissue around the coronary artery is part of the EAT and was defined as the attenuation coefficient of the fat tissue voxels within a distance from the coronary arterial wall equal to the corresponding vessel diameter. The attenuation coefficient was calculated separately for the right coronary artery (RCA), left anterior descending artery (LAD), and left circumflex artery (LCX) using the coronary artery analysis module [18].

The cardiac mass was determined using the syngo.via cardiac analysis module, which automatically traces the endocardial and epicardial borders to generate the volume of the heart. The myocardial volumes were converted to left ventricle mass (M) by assuming a constant tissue density of myocardium (1.05 g/cc). CTV/M was calculated by dividing the luminal volume of the coronary artery by the myocardial mass of the heart [9, 19]. The luminal volume of the coronary artery was extracted by a deep learning calculation method on the workstation. In this study, the right coronary artery was selected for coronary lumen volume analysis given its simplicity and thick diameter in the Chinese population.

Follow-up included follow-up data and results obtained from January 2019 and June 2022 in the HIS system as a follow-up of disease and care, clinic visits, and telephone interviews. The endpoints were patient readmission (defined as a hospital admission for which HF was the primary reason and requiring either diuretic, inotropic, or intravenous nitrate therapy). All statuses were reviewed by 2 independent investigators who used previously described criteria.

Data were analysed using R software (Version 4.0). Continuous variables are reported as the mean ± SD or median (interquartile range, IQR) and were compared using independent samples t tests and one-way analysis of variance or Wilcoxon signed-rank and Kruskal‒Wallis tests. Categorical variables are presented as absolute values and percentages. The correlation between healthy controls and HF patients was analysed using the Pearson correlation. Inter- and intrao bserver agreement for the CTEAT, CTMASS, CTV/M, CTFFR and CTPCAT were performed using Bland–Altman analysis.

For survival analysis, survival and proportional hazards assumptions were estimated by the Kaplan‒Meier method, and any differences were evaluated with the stratified log-rank test. The optimal cut-off point was identified using the maximally selected rank statistic (maxstat) [20], which is a ranking statistic for maximum selection. Odds ratios (ORs) and 95% confidence intervals (CIs) were determined from univariate and multivariate logistic regression analyses to determine any factors associated with the endpoint. A favourable patient prognosis was taken as the dependent variable, and the statistically significant factors were included in the logistic regression model as independent variables for regression analysis. Three risk models were created based on the best traditional parameter variables (model 1), the CCTA-driven parameters (model 2), and the combination of the two sets of parameters (model 3). Receiver operating characteristic (ROC) curve analysis was performed to evaluate the prognostic accuracy of CTEAT, CTMASS, CTV/M, CTFFR and CTPCT for the endpoint.