The study was approved by the Institutional Review Board of Mackay Memorial Hospital, Taipei, Taiwan. All participants signed written informed consent prior to examinations. Data were analyzed anonymously. From 2005 to 2012, a total of 719
consecutive subjects underwent cardiovascular health survey at our center that included a non-contrast enhanced computed tomography (CT) scan of the heart for coronary calcium scoring. A subset of 704 participants also had a liver ultrasound scan were eligible for the inclusion of the present study. Ultrasonography was performed using Philips EPIQ Ultrasound Machine. The images were interpreted by board certified gastroenterologists who were unaware of the clinical or laboratory data of the participants. Fatty liver was assessed, based on the presence of increased hepatic echogenicity making it distinguishable from the renal parenchyma of liver. Mild fatty liver was assessed as the minor increase in liver echogenicity. In moderate fatty liver, there were visual images associated with intrahepatic vessels, the slightly damaged diaphragm and the existence of increased liver organ echogenicity. Severe fatty liver was defined as the significant increase in hepatic echogenicity, poor penetration of posterior segment from the right lobe of the liver, poor or any visual images from the hepatic vessels and diaphragm. We defined NAFLD as fatty liver in individuals whose alcohol use disorders identification test (AUDIT) score was less than 8. Baseline demographics and medical history were obtained along with a detailed physical exam. Structured questionnaires were used to quantify self-reported alcohol consumption, smoking and physical activity.
Baseline anthropometrics and metabolic syndrome
A variety of anthropometric measures including
height, weight, waist and hip circumferences were obtained. Resting blood pressures were measured by medical staff using a standardized sphygmomanometer. Anthropometric measures collected were height, weight, body mass index (BMI), waist and hip circumference. Standardized blood pressures were measured at rest by medical staff blinded to the other test results. Total body fat mass was measured by bioelectrical impedance using a Tanita-305 foot-to-foot body-fat analyzer (Tanita Corp., Tokyo, Japan). The definition of metabolic syndrome used a waist circumference cut-off of ≥ 90 cm and 80 cm for Taiwanese men and women, respectively. Additional criteria were: systolic blood pressure ≥130 mmHg or diastolic blood pressure ≥85 mmHg, triglyceride level ≥150 mg/dL, fasting blood sugar level ≥100 mg/dL, and HDL ≥40 and 50 mg/dL in men and women, respectively. The metabolic score therefore ranged from 0 to 5.
The presence of metabolic syndrome (MetS) was defined as a metabolic score of 3 or more.
Pericardial (PCF) and thoracic periaortic adipose tissue volume (TAT)
Pericardial (PCF) and thoracic peri-aortic adipose tissue (TAT) volumes were quantified from the ECG-gated non-enhanced cardiac CT images using a dedicated workstation (Aquarius 3D Workstation, TeraRecon, San Mateo, CA, USA). The semi-automatic segmentation technique was developed for quantification of adipose tissue volumes. We traced the region of interest manually and defined adipose tissue as pixels within a window of -195 HU to -45 HU and a window centre of -120 HU. PCF was defined as all adipose tissue located within the pericardial sac. TAT tissue was defined as all adipose tissue surrounding the thoracic aorta extending 67.5 mm caudally from the level of the bifurcation of pulmonary arteries. This approach has previously been validated [7-8]. The intra-observer and inter-observer coefficient of variation were 4.27%, 4.87% and 6.58%, 6.81% for PCF and TAT .
All the analyses were performed by using SPSS 15 (SPSS Inc., Chicago, IL). The characteristics of study subjects were expressed either as mean±SD or frequency with percentage. Study subjects were divided into three groups according to their degree of fatty liver diagnosis: normal, mild, moderate and severe. Linear contrast in general linear model was used to examine the trend of each continuous variable across groups; Mantel-Haenszel chi-squared test was used for categorical variables. Each P value for linear trend was reported.
Concerning with the ordinal nature of the fatty liver diagnosis, ordinal logistic regression was applied. The results of ordinal logistic regression are presented as the odds ratio (OR) and 95% confidence interval (CI) of being in a more severe fatty liver level for 1-unit change in serum parameters or for the presence or absence of medical history/life style variables.
The association of biomarkers－PCF and TAT－with fatty liver was assessed in different adjustment logistic models. In addition to these two biomarkers, models also included (1) age and gender; (2) age, gender, and established risk factors (3) age, gender, established risk factors, and life styles. Established risk factors were systolic blood pressure (SBP), fasting glucose, triglyceride, high-density cholesterol (HDL), cholesterol, eGFR, hypertension, diabetes, and hyperlipidemia. Life style factors contained regular exercise (yes vs. no), alcohol consumption (ever vs. never), and smoking status (ever vs. never). Each anthropometric factor－BMI, body fat, or waist circumstance－was further adjusted in Model 4, separately.
To identify the incremental values of PCF and TAT for the diagnosis of fatty liver beyond metabolic syndrome, likelihood ratio test was performed. Areas under ROC curve (AUC) and 95% CIs of each biomarker were reported to discriminate the prediction for fatty liver severity (moderate and severe vs. normal/mild) from metabolic syndrome.