The study protocol was approved by the Institutional Review Board Committee of Kaohsiung Veterans General Hospital, Kaohsiung, Taiwan as No. VGHKS 19-CT6-02. The requirement for informed consent was waived due to retrospective study design by the ethics committee of Kaohsiung Veterans General Hospital. All methods were performed in accordance with the relevant guidelines and regulations. In this retrospective study, we reviewed the records of 972 subjects who underwent routine health checkups, abdominal sonogram and coronary CT angiography examination at the same time in Kaohsiung Veterans General Hospital from January 1, 2018 to December 31, 2018. The inclusion criteria were as follows: (1) all subjects underwent abdominal sonogram and coronary CT angiography examination at the same time; (2) all subjects are asymptomatic. 155 subjects with at least one of the following: (1) a history of ≥ 10 g of daily alcohol consumption; (2) liver cirrhosis (defined by ultrasonographic criteria); (3) chronic hepatitis B or C (defined by history, serum hepatitis B surface antigen, and anti-hepatitis C antibodies) were excluded to meet the diagnostic criteria of NAFLD according to the National Health and Nutrition Examination Survey III criteria [13]. Of the remaining 817 subjects were enrolled into the study population for further analysis show in Fig. 1. Obtained subjects’ demographics, cardiovascular risk factors and anthropometric measurements were recorded from referral visit information in the electronic medical records (EMR) system, including age, gender, body mass index (BMI), hypertension, systolic blood pressure, hypertension medication, diabetes mellitus, current smoking habits, the amount of pack-year, waist circumstance, body-fat percentage, and alcohol intake (number, frequency, and alcohol percentage of drinks per week according to the National Institute on Alcohol Abuse and Alcoholism) were recorded for all participants. Anthropometric measurements such as BMI and body-fat percentage were measured using the electric impedance method analyzer (XSCAN PLUS II; Jawon Medical, Gyeongsan-si, South Korea) with the patients minimally clothed and wearing no socks as the previous study described [14]. Complete biochemical and blood examinations including blood levels of low density lipoprotein-cholesterol (LDL-C), high density lipoprotein-cholesterol (HDL-C), triglyceride (TG), total cholesterol, estimated glomerular filtration rate (eGFR), hemoglobin A1c (HbA1c) concentration were also recorded at the same time. Framingham risk score (FRS) percentage was calculated based on the six coronary risk factors including age, gender, total cholesterol, HDL-C, systolic blood pressure, and smoking habits for conventional cardiovascular risk stratification[15].

The diagnosis of severity of NAFLD was based on the ultrasonographic findings established according to the practice guideline of the American Gastroenterology Association [16]. All ultrasonographic studies were performed by two imaging units (GE Logic E9, GE Healthcare, Wauwatosa, WI, USA; Acuson S2000, Siemens, Germany) with a 3.5- to 5-MHz convex probe. The severity of NAFLD was divided into the normal, mild, moderate, and severe degree based on hepatic ultrasonography. Mild fatty liver was defined as slightly increased liver parenchyma echogenicity compared to kidney/spleen with normal visualization of portal vein wall and diaphragm. Moderate fatty liver was defined as the moderate increase of liver parenchyma echogenicity compared to kidney/spleen with mild dimness of portal wall and diaphragm. Severe fatty liver was defined as the high increase of liver parenchyma echogenicity compared to kidney/spleen with poor visualization of portal wall and diaphragm due to elevation of echogenicity [17]. All images were retrospectively reviewed by one single expert radiologist with 10 year experience in ultrasonography. The expert was blinded to the clinical and laboratory data of subjects and unaware of the previously study reported.

All enrolled study subjects underwent initial CAC scans and sequential coronary CT angiography examinations at the same time with 256-slice CT scanner (Revolution CT, GE Healthcare, Milwaukee, USA). We administered oral beta-blockers (metoprolol) if the heart rate exceeded 65 beats per minute an hour before the exams began to start. All subjects received 0.8 mg of sublingual nitroglycerin shortly prior to the CAC scan. The images were analyzed using a dedicated workstation (AW Server, version 3.2, GE Healthcare). All coronary CT angiography examinations were read by two experienced cardiovascular radiologists. Imaging findings of coronary artery stenosis including luminal stenosis severity, plaque composition, and vulnerable plaque(s) characteristics were analyzed according to Coronary Artery Diseases Reporting and Data System (CAD-RADS) classification by 2016 SCCT guideline, including twenty-five coronary segments on the basis of the percentage of cross-sectional area stenosis [18] shown in Additional file 1: Table S1. For CAC score measurements, noncontrast ECG-gated cardiac CT scans were performed based on the Agatston method by using a commercially available CT workstation (AW Server, version 3.2, GE Healthcare). CAC scores were categorized as 4 categories according to the degree of calcification (0–10: minimal; 10–100: mild; 100–400: moderate; > 400: severe) [19‐21]. In this study cohort, primary endpoints including CAC score ≧ 100, CAC score ≧ 400, CAD-RADS ≧ 3 and presence of vulnerable plaque(s) were assessed for severity of NAFLD associated with subclinical coronary atherosclerosis. The definition of CAD-RADS ≥ 3 was that presence of significant luminal stenosis 50–69%% in any one of the triple coronary trees. There are four main vulnerable plaque features on coronary CT angiography: (1) positive remodeling (2) low attenuation plaque (3) napkin-ring sign (4) spotty calcium. Presence of at least two high-risk plaque feature (s) can be indicated with the definition of vulnerable plaque (s) (Additional file 1: Table S1).

Subclinical coronary atherosclerosis was defined as the presence of atherosclerotic plaque or coronary artery calcification score (CACS) ≥ 1 in asymptotic subjects.

We included 817 subjects (mean age 54.20 ± 10.02 years, 60.20% male) with detailed baseline characteristics and anthropometric and laboratory measurements demonstrated. Among 817 study subjects, 600 (73.43%) subjects have different degrees of NAFLD. The baseline characteristics, anthropometric and laboratory measurements of the study cohort subjects (N = 817) classified into 4 groups according to the NAFLD severity are shown in Table 1. There were statistically significant differences across the four groups for the in the clinical characteristics, anthropometric and laboratory measurements and coronary artery finding demonstrated by one-way ANOVA, except for current smoking habits, the amount of pack-year, body-fat percentage, eGFR and LDL-C level. In terms of Bonferroni post-hoc tests (inter-group comparison) are also shown in Table 1.

As shown in Fig. 2, subclinical coronary atherosclerosis in term of CAC ≧ 100, CAC ≧ 400, CAD-RADS ≧ 3 and presence of vulnerable plaques ( +) significantly increased as the severity of NAFLD increased by using the general linear model for trend analysis (CAC ≧ 100, p for trend < 0.0001; CAC ≧ 400, p for trend = 0.003;CAD-RADS ≧ 3, p = 0.015; presence of vulnerable plaques, p for trend = 0.004.

Figure 2 shows that the primary endpoints about subclinical coronary atherosclerosis significantly increased as the severity of NAFLD increased by using the general linear model for trend analysis. In addition, post-hoc analysis revealed no significant difference in group 1(mild NAFLD) versus group 2 (moderate NAFLD) and group 0 (normal) versus group 1 (mild NAFLD) in terms of subclinical coronary atherosclerosis. Therefore, in the binary logistic regression, the reference group (including normal, mild NAFLD, moderate NAFLD groups) was chosen as the reference to investigate the relationship between the severity of NAFLD and subclinical coronary atherosclerosis. Table 2 shows the multivariate logistic regression analysis to determine the predictors of subclinical coronary atherosclerosis in term of CAC ≧ 100, CAC ≧ 400, CAD-RADS ≧ 3 and presence of vulnerable plaque(s) with adjustment for FRS score and body-fat percentage. For investigation the association with CAC ≧ 100, severe degree NAFLD is the independent predictor of CAC ≧ 100 after adjusting for FRS score and body-fat percentage (odds ratio of 2.946, 95% CI 1.017–8.530, p = 0.046). For investigation the association with CAC ≧ 400, severe degree NAFLD is the independent predictor of CAC ≧ 100 after adjusting for FRS score and body-fat percentage (odds ratio of 4.402, 95% CI 1.149–16.862, p = 0.031). For investigation the association with CAD-RADS ≧ 3, severe degree NAFLD is the independent predictor of CAD-RADS ≧ 3 after adjusting for FRS score and body-fat percentage (odds ratio of 4.091, 95% CI 1.396–11.993, p = 0.010). For investigation the association with presence of vulnerable plaque(s), severe degree NAFLD is the independent predictor of presence of vulnerable plaque(s) after adjusting for FRS score and body-fat percentage (odds ratio of 2.906, 95% CI 1.107–7.624, p = 0.030).