A total of 2,318 subjects were enrolled in the phase-I of the Indian Atherosclerosis Research Study (IARS) (10) between 2004–2006, which is a prospective study designed according to the principles and the guidelines of World Medical Association Declaration of Helsinki and the Indian Council of Medical Research and approved by Thrombosis Research Institute Ethics committee. Signed informed consent was obtained from all the participants who were enrolled, and 240 age- and gender-matched subjects (120 unaffected and 120 affected) were selected for this study. All the subjects were ≤45 years of age for males and ≤50 years for females at the onset of the disease. The individuals with an abnormal electrocardiogram (ECG), positive angiogram showing >70% stenosis in any one of the coronary arteries or >50% in two or more arteries and individuals posted for coronary artery bypass graft (CABG) surgery or percutaneous coronary intervention were considered as CAD subjects. The unaffected subjects were all healthy volunteers without clinical signs of CAD, as confirmed by ECG. None of the participants enrolled in the study had inflammatory disorders, concomitant infections or any other major illness such as cancer, liver or renal disease.

Periodic telephonic follow-up of all the study participants was undertaken to obtain an update on the cardiac health status, and five rounds of follow-up for an average time of 18 months were completed. Endpoints included non-fatal events (re-do percutaneous transluminal coronary angioplasty/repeat CABG/heart attack or stroke) and fatal events (fatal heart attack/fatal stroke/sudden death or cardiac arrest). This information was verified through either paper trail or hospital records, wherever available.

All the participants were assessed for the presence of the risk factors by personal interviews, which included demographics (age, gender, occupation and socio-economic status), lifestyle, health characteristics (diet, waist circumference, smoking status and use of alcohol), and medical history [hypertension and diabetes, systolic and diastolic blood pressure, LDL and HDL-cholesterol (HDL-c), and triglycerides].

Venous blood samples (20 ml) were collected after overnight fasting (12 h). The samples were centrifuged at 2,000 × g for 20 min at 25°C, following which 1 ml aliquots of serum and plasma were stored at −80°C. Total cholesterol and triglycerides were measured using reagents and standards from Randox Laboratories (Antrim, UK) in a Cobas-Fara II Clinical Chemistry auto-analyzer (Hoffman La Roche Ltd., Basel, Switzerland). For the estimation of HDL-c, the non-HDL fractions were precipitated with a mixture of 2.4 mmol/l phosphotungstic acid and 39 mmol/l magnesium chloride (Bayer Diagnostics, Baroda, Gujarat, India) prior to estimation. The LDL concentration was calculated using the Friedewald formula. The inter-assay coefficient of variation for commercial controls and normal human pooled serum/plasma (NHP) was 4.9–7.0% for total cholesterol, 6.1–7.7% for triglycerides, and 7.1–12.2% for HDL-c.

All the plasma or serum biomarkers were measured by ELISA (enzyme-linked immunosorbent assay) or the immunoturbidometric method. Plasma GGT and high-sensitivity C-reactive protein (hsCRP) concentrations were measured using a Flex reagent cartridge in Dimension XpandPlus Integrated Chemistry system (Siemens, Germany). Lipoprotein (a) [lp(a)] (Randox Laboratories) was measured in the Cobas-Fara II Clinical Chemistry auto-analyzer. Plasma levels of interleukin-6 (IL-6), cystatin C (R&D systems, Minneapolis, MN, USA), neopterin (IBL International, Atlanta, GA, USA), myeloperoxidase (MPO) (Mercodia, Uppasala, Sweden) and tumor necrosis factor-like weak inducer of apoptosis (TWEAK) (eBiosciences, Vienna, Austria) were measured using ELISA kits. The secretory phospholipase A2 (sPLA2) was measured by a sandwich immunometric assay (Cayman Corporation, Ann Arbor, MI, USA). The inter-assay coefficient of variation of in-house controls (NHP) was 4.40% for GGT, 5.2% for lp(a), 4.3% for IL-6, 5.37% for sPLA2, 7.85% for hsCRP, 5.99 to 11.8% for neopterin, 11.8% for cystatin C, 9.3% for myeloperoxidase, and 7% for TWEAK.

The variations in the baseline characteristics between the unaffected and affected subjects were studied using the Student's t-test and univariate analysis of variance for quantitative variables. The quantitative parameters were adjusted for age, gender, hypertension, diabetes mellitus and smoking. χ² test was used to assess the association of discrete outcomes with premature CAD. Results are represented as mean ± standard error for quantitative variables, whereas the discrete outcomes were represented as frequencies. The Shapiro-Wilk and Kolmogorove-Smirnov test were used to assess the normality of the data. The data with skewed distribution were natural log-transformed. The association of GGT with lipids, inflammatory and oxidative stress markers was evaluated by Pearson's correlation for quantitative variables and partial correlation was used for conventional risk factors (CRFs; which were age, gender, hypertension diabetes mellitus and smoking) adjusted analysis. The association of GGT with discrete outcomes was evaluated using χ² test.

In order to review the association between higher levels of GGT and incidence of premature CAD, GGT levels were categorized in to tertiles. The odds ratios (OR) with 95% confidence interval (CI) were obtained by comparing top tertiles, considering the first tertile as the reference category using logistic regression analysis. The logistic regression analysis was also used to build different models with multiple markers having GGT in common. The discriminative ability of a significant model with the highest OR was evaluated using receiver operating characteristic analysis. The significance of improvement in the area under the curve (AUC) between CRF and GGT + CRF models was assessed using DeLong's test. A two-tailed P-value <0.05% was considered to indicate a statistically significant difference. All the above statistical analyses were carried out using statistical package for social sciences (SPSS) version 17.0 (SPSS, Inc., Chicago, IL, USA).

The clinical characteristics of the 240 age- and gender-matched study subjects (120 affected and 120 unaffected) are provided in Table I. The CAD-affected subjects showed a high frequency of hypertension and diabetes, and low levels of HDL-c compared to the unaffected subjects. The frequency of smokers was also significantly higher in the CAD-affected group. The affected subjects showed a lower level of LDL compared to the unaffected, possibly due to statin usage in this group. Six out of 9 protein markers assayed in the present study were significantly different between the affected and unaffected subjects. As shown in Table I, the mean GGT levels were significantly higher in CAD-affected subjects in comparison with the unaffected subjects (34.48±1.00 vs. 30.94±0.90 U/l; P=0.009) and similar results were also obtained for lp(a) (27.20±2.34 vs. 18.70±1.55 mg/dl; P=0.019).

The GGT levels showed a significant moderate-positive correlation with the oxidative stress marker neopterin, and remained borderline significant following adjustment for the age, gender, hypertension, diabetes mellitus and smoking (Table II). GGT was also positively correlated to triglyceride levels and waist circumference, and remained significant following adjustment with age, gender, hypertension, diabetes mellitus and smoking (all the significant correlation coefficients were <0.3).

Cross-sectional correlation was carried out to understand the association between GGT and binary outcomes. The present analysis did not observe any association of hypertension, diabetes and alcohol habit with GGT levels (Table III). The number of smokers in the second and third tertile of GGT was higher in comparison to the first, and there was a positive association between GGT levels and smoking.

In comparison to the low GGT level of <22.59 U/l reference group, the high GGT levels of ≥32.21 U/l (third tertile) showed an OR of 1.909 (95% CI, 1.036–3.517; P=0.038). Following adjustment to the GGT-associated risk factors and CAD risk factors, the OR increased to 2.104 (95% CI, 1.063–4.165; P=0.033), i.e., the subjects with GGT levels of ≥32.21 U/l had a 2.11-fold higher risk of premature CAD in comparison to <22.59 U/l. The intermediate GGT levels (second tertile) group did not show any association with premature CAD prior and subsequent to adjustment to the conventional risk factors (Table IV).

The logistic regression model was used to assess the risk predictive ability of GGT alone and along with lipid, inflammatory, oxidative stress markers and conventional risk factors. GGT alone was found to have an OR of 4.94 (95% CI, 1.015–24.095; P=0.048), which improved to 5.208 (95% CI, 1.018–24.624; P=0.048) upon adjusting for the conventional risk factors. The OR of this model further improved to 5.64 (95% CI, 1.004–31.642; P=0.049) upon addition of lp(a). The above combination of GGT and lp(a) exhibited a higher odds ratio of 7.492 (95% CI, 1.221–45.979; P=0.030) (Table V) following adjustment for CRFs.

The addition of MPO to the GGT was significant with a reduction in the OR of GGT from 4.94 to 3.48; however, the significance of the model was lost following adjustment with CRFs (Table V). Other inflammatory and oxidative stress markers did not add any significant value to the predictive ability of the GGT.

The C-statistic analysis was performed to assess the discriminative ability of the models built using logistic regression (Fig. 1). The discriminative ability of GGT alone was tested and compared with the other models. The GGT alone significantly classified the 61.5% population correctly. However, conventional risk factors used in the model building exhibited an AUC of 0.647 (95% CI, 0.577–0.717; P<0.001). The GGT and lp(a) together increased the AUC to 0.654 (95% CI, 0.576–0.732; P<0.001). A combined model of GGT and lp(a) plus conventional risk factors showed the best performance with an AUC of 0.711 (95% CI, 0.642–0.799; P<0.001). The DeLong's test indicated that the improvement of AUC from CRF to biomarkers [GGT + lp(a)] adjusted to CRFs was significant (P<0.001). All the models built were tested for best-fit using the Hosmer-Lemeshow test.