This was a prospective observational study. Consecutive patients hospitalized at the Second Affiliated Hospital of Soochow University (Suzhou, China) for the diagnosis and management of CAD were recruited. To confirm the diagnosis of CAD, only patients who had undergone coronary angiography and were found to have ≥50% stenosis in at least one coronary artery were deemed eligible. Patients with end-stage renal failure, severe hepatic insufficiency, current microbial infections or malignant tumor were excluded. Details about the study protocol were explained to the eligible patients and informed consent was obtained. The study protocol was approved by the Medical Ethics Committee of the Second Affiliated Hospital of Soochow University.

Samples of venous blood were collected following overnight fasting, and stored at −80°C prior to analyses. Serum homocysteine level was measured by an enzymatic conversion method (Formosa Biomedical Technology Corporation, Taiwan); serum adropin level with a commercial enzyme-linked immunosorbent assay (ELISA) kit (JRDUN Biotechnology Co., Ltd., Shanghai, China); high sensitivity C-reactive protein (hsCRP) level with a high-sensitivity ELISA kit (Orion Diagnostica Oy, Espoo, Finland), and fasting plasma glucose with the hexokinase method (AU5400 High-Volume Chemistry Immuno Analyzer; Olympus Corporation, Tokyo, Japan). Serum creatinine level and serum lipid profiles including triglyceride, total cholesterol, low-density lipoprotein cholesterol and high-density lipoprotein cholesterol were assessed using standard methods. In brief, serum creatinine level was measured using the muscle amino acid oxidase method. Serum total cholesterol and triglyceride levels were measured by automated enzymatic procedures. The low-density lipoprotein cholesterol and high-density lipoprotein cholesterol levels were determined after separating the lipoprotein fractions from fresh fasting sera by sequential ultracentrifugation, using the AU5400 analyzer.

The demographic and clinical characteristics of the recruited patients were collected from hospital case records. These included age, gender, cigarette smoking status, hypertension, diabetes mellitus, and principal diagnosis at the index admission. Echocardiograms were carried out to determine the left ventricular ejection fraction prior to coronary angiography. Patients' height (m) and weight (kg) in light clothing were measured, and the body mass index (kg/m²) was calculated.

Coronary angiography was performed using standard techniques through radial or femoral approaches. Angiographic analysis was carried out by two experienced interventional cardiologists who were blinded to the study. The number of diseased coronary arteries (luminal diameter narrowing ≥50%) was recorded, and patients with stenosis of the left main coronary artery ≥50% were considered to have two-vessel disease. SYNTAX scores of the study patients were calculated according to the guidelines of the SYNTAX Steering Committee and Boston Scientific Corporation (http://www.syntaxscore.com/).

Exploratory analyses involving quantitative and qualitative data were performed with independent-sample t-tests (or the nonparametric Wilcoxon-Mann-Whitney tests if the data were non-normally distributed) and Chi-square tests, respectively. The association between homocysteine level, adropin level and SYNTAX score were assessed by pairwise Pearson correlation. In the following confirmatory analyses, a generalized structural equation model (gSEM) (20) was built with multiple risk factors and confounders in order to ascertain their effects on hyperhomocysteinemia (HyperHCY) and SYNTAX scores (SYNTAX). Patients were classified as suffering from hyperhomocysteinemia if their serum homocysteine level was >15 µmol/l (21,22). The specific variables considered in the gSEM included demographic, clinical and angiographic characteristics. The details are depicted in Table I.

The proposed gSEM model, unlike the conventional regression, can accommodate multiple outcomes and handle more complex inter-relationships among the variables. It is a system of related models combined and estimated within a single framework, which is able to estimate the direct, indirect and total effects of the variables on the outcomes. As there were two outcomes of interest, the proposed gSEM model was constructed with two equations. SYNTAX was dealt with by a linear regression, and HyperHCY was analyzed with a binomial distribution and logit link in view of its binary nature. The data were exported to Stata/MP version 13 (Stata Corporation, College Station, TX, USA) for analysis, and all statistical tests were conducted with a 5% level of significance.

Between July 2012 and May 2013, 170 consecutive patients with a mean serum homocysteine level of 15.9±8.3 µmol/l, were successfully recruited into the study. The patients' demographic, clinical and angiographic characteristics are presented in Table I. Among the recruited patients, 76 (44.7%) had suffered from hyperhomocysteinemia (serum homocysteine level >15 µmol/l). When compared with patients without the condition, they were found to have a lower serum adropin level, but higher body mass index, SYNTAX score and creatinine level. In addition, patients with hyperhomocysteinemia were more likely to have 3-vessel disease, and less likely to have single vessel disease. No significant differences in age, gender, smoking status, hypertension, diabetes, left ventricular ejection fraction, fasting plasma glucose, cholesterol, triglyceride, low density lipoprotein cholesterol, high density lipoprotein cholesterol, hsCRP and principal clinical presentations were observed between the two groups.

The recruited patients reported a mean SYNTAX score of 21.5±11.2 (range, 4–57). The serum homocysteine level was significantly negatively correlated with the serum adropin level (Pearson correlation r=−0.169, P=0.028). Although the serum homocysteine level was not significantly correlated with SYNTAX score (r=0.124, P=0.108), hyperhomocysteinemia was (r=0.238, P=0.002). The SYNTAX score was in turn negatively correlated with the adropin level (r=−0.181, P=0.018). There is thus evidence suggesting that serum adropin, while negatively associated with serum homocysteine, could predict coronary atherosclerosis quantified with SYNTAX score. Further pairwise correlation analyses revealed that SYNTAX score was significantly associated with age (r=0.183, P=0.017) and serum creatinine (r=0.364, P<0.001).

In view of the relatively large number of predictors considered in the confirmatory analysis with gSEM, a backward elimination routine (with P≥0.05 as the criterion for variable elimination) was implemented to identify which demographic, clinical and angiographic variables were significantly associated with HyperHCY and SYNTAX, while considering the effect of serum adropin level. Extra consideration was taken when a large effect size was found, despite its statistical non-significance. The standard errors of the estimated effects were corrected with the robust statistical procedure. This ensured that the constructed gSEM was not sensitive to invalid modeling assumptions, particularly when there were potential outliers that could not be well accommodated with any parametric framework.

The final gSEM identified adropin and creatinine as the significant predictors of HyperHCY, and indicated that adropin and HyperHCY affected the SYNTAX score (Table II). As revealed in the HyperHCY sub-model, adropin had a negative impact [adjusted odds ratio (AOR): 0.95, 95% confidence interval (CI): 0.93 to 0.98; P=0.002) on HyperHCY and this conformed with the earlier pairwise correlation analysis. However, a higher level of creatinine (AOR: 1.02, 95% CI: 1.00 to 1.04; P=0.046) was significantly associated with HyperHCY. It is worthy of note that hypertension was retained in the sub-model in view of its large effect size. When other variables were comparable, CAD patients suffering from hypertension were 1.87-fold more likely to have hyperhomocysteinemia when compared with those without hypertension, although the effect was non-significant.

Adropin and HyperHCY were significantly associated with SYNTAX when adjusting for the effect of age, as revealed in the SYNTAX sub-model (Table II). All other variables being comparable, a unit increase in serum adropin level (pg/ml) would reduce the SYNTAX score by 0.04 units (95% CI: −0.06 to −0.01; P=0.019). Patients with hyperhomocysteinemia had a significantly higher average SYNTAX score of 4.71 units (95% CI: 1.39 to 8.03; P=0.005), when compared with those not suffering from the condition. Finally, as the CAD patients aged their SYNTAX score was expected to increase (coefficient: 0.19, 95% CI: 0.04 to 0.34; P=0.012). The analyses with gSEM confirmed that the study hypotheses were valid. While serum adropin was significantly associated with hyperhomocysteinemia, both low adropin level and hyperhomocysteinemia were significantly associated with coronary atherosclerosis.