Study population

All consecutive patients with chronic coronary syndromes (CCSs) scheduled for elective coronary angiography at the Department of Cardiology (Medical University of Silesia in Katowice, Poland) were screened between 2012 and 2013 using the study criteria.⁴ Finally, 150 patients were prospectively enrolled into the study group.

The main exclusion criteria were similar to those described previously,⁴ and the most important ones included acute coronary syndrome, heart failure, chronic inflammatory diseases, neoplastic diseases, severe kidney or liver failure, secondary causes of obesity, and an unintentional weight loss of 10% in the prior 3 months or malnutrition.

Pharmacotherapy in all patients was standardized according to the appropriate European Society of Cardiology guidelines.^10-13 The baseline study visit (2012–2013) comprised a clinical assessment and measurement of anthropometric parameters, as well as evaluation of ultrasound vascular indexes and parameters of fat depots as described earlier.⁴

Afterwards, all the patients were invited for prospective follow‑up visits in 2020. A total of 143 patients signed an informed consent and agreed to participate in the current study. The prospective follow‑up visit included an evaluation of the medical history as well as the clinical, anthropometric, and laboratory parameters.

The study aimed to assess the predictive value of ultrasound indexes and anthropometric parameters for cardiometabolic risk in patients with a high cardiovascular risk profile in an 8‑year follow‑up. The primary outcome was a new diagnosis of DM and / or insulin resistance (IR). The secondary outcome was all‑cause mortality.

The study protocol complied with the Declaration of Helsinki and was approved by the local Medical University of Silesia Ethics Committee (KNW/0022/KB1/127/I/13/14 and KNW/0022/KB1/77/17).

Clinical characteristics and anthropometric measurements

Hypertension was defined according to blood pressure (BP) levels (systolic BP ≥140 mm Hg and / or diastolic BP ≥90 mm Hg) or a history of hypertension and current antihypertensive treatment.¹⁴ The presence of hyperlipidemia was established based on abnormal plasma lipid levels (total cholesterol [TC] >190 mg/dl, low‑density lipoprotein cholesterol [LDL‑C] >115 mg/dl, triglycerides [TG] >150 mg/dl, high‑density lipoprotein cholesterol [HDL‑C] <⁠40 mg/dl in men and <⁠50 mg/dl in women) or prior diagnosis and / or current treatment of hyperlipidemia.¹⁴ The diagnoses of the analyzed cardiovascular diseases and risk factors were defined according to the European Society of Cardiology guidelines¹⁵ and identified based on a prior diagnosis and / or current treatment of these diseases. DM was determined based on fasting plasma glucose levels greater than or equal to 126 mg/dl and / or glycated hemoglobin (HbA_1c) concentration greater than or equal to 6.5% or current antidiabetic treatment.¹⁶

Overweight and obesity were diagnosed based on the body mass index (BMI; calculated as body mass [kg] / height [m²]) as normal weight (18.5–24.9 kg/m²), overweight (25.0–29.9 kg/m²), and obesity (≥30.0 kg/m²), which was further divided into class I (30.0–34.9 kg/m²), class II (35.0–39.9 kg/m²), and class III (≥40.0 kg/m²) obesity.

Waist circumference (WC; midpoint between the lowest rib and the iliac crest) and hip circumference were assessed using a measuring tape at the end of expiration. Increased WC was defined as per the World Health Organization (WHO; ≥102 cm in men and ≥88 cm in women) and the International Diabetes Federation (IDF; ≥94 cm in men and ≥80 cm in women) criteria. Additional measurements included: neck circumference, thigh circumference, and upper limb circumference (evaluated together as total circumference [TCirc]). Moreover, we measured the following skinfold thicknesses: chest, abdominal, and thigh (evaluated together as total skinfold thickness) as previously described.⁴

We used the bioelectric impedance analysis method (Bodystat 1500, Bodystat, Douglas, Isle of Man) to measure the patients’ body composition, including body fat percentage, similarly to the method used before.⁴ All measurements were performed according to the manufacturer’s manual. The intraobserver variability of measurements in the same patient was below 1%.

Sample collection and routine laboratory tests

Fasting blood samples (15 ml) were obtained at baseline and after 8 years of follow‑up. At both time points, the blood was drawn from an antecubital vein between 8 and 10 am after overnight fasting. The samples were processed 30 to 60 minutes after collection. Routine blood tests, including the evaluation of complete blood count and lipid profile (TC, LDL‑C, HDL‑C, and TG), were performed using automated laboratory techniques. HbA_1c levels were measured with a turbidimetric inhibition immunoassay.

Ultrasound indexes

The ultrasound indexes were evaluated as described before.^4,5,17 Briefly, all the ultrasound examinations were performed by a single experienced physician using constant and standardized settings. After the enrollment, single images of an individual ultrasound index of interest were randomly analyzed by 1 observer blinded to the patients’ data. Additionally, a random number of 40 scans were analyzed again by a second experienced observer to determine the intra- and interobserver variability and coefficients of variation.

Before electrocardiogram‑gated ultrasound measurements of both carotid arteries, all the patients rested in a supine position for at least 5 minutes. The examination was performed using a high‑resolution ultrasound device (GE Vivid 9, Milwaukee, Wisconsin, United States) with a linear transducer (9–12 MHz). The common carotid artery (CCA) intima‑media thickness (IMT) was measured on a 10‑mm segment starting 5 mm proximally to the carotid bulb, as stated in the Mannheim Consensus guidelines,⁶ using a semiautomated GE software (intra- and interobserver coefficients of variation, 2.6% and 3.1%, respectively). The carotid IMT was measured on both sides, and average values were calculated. As defined before,⁵ extra‑media thickness (EMT) was the distance between the carotid media‑adventitia border and the jugular wall lumen area, averaged from both CCAs. The EMT evaluation was performed after zooming the interface between the wall of the distal segment of the CCA and the neighboring jugular vein (intra- and interobserver coefficients of variation, 3.3% and 3.9%, respectively).

Epicardial and pericardial adipose tissue was quantified using transthoracic echocardiography with a 1.5–4.5 MHz transducer according to the established method (GE Vivid 9, GE Vingmed Ultrasound AS, Horten, Norway).^5,18 Epicardial fat thickness (EFT) was defined as the distance between the epicardium and the visceral layer of the pericardium. Pericardial fat thickness (PFT) was measured as the distance between the visceral and parietal pericardial layers. The maximum values were measured during the end‑diastole phase in 5 consecutive heart cycles and mean values were obtained (intra- and interobserver coefficients of variation, 3.3% and 3.9% for EFT and 3.5% and 4% for PFT, respectively). As described before,^4,5 visceral adipose tissue measurements were performed using a 3.5 MHz transducer, and for subcutaneous fat (SF; to determine the maximal SF layer) assessment, using a 7.5 MHz transducer placed vertically on the skin as lightly as possible to prevent compression and to perform breath‑hold assessments of fat layers. Intra‑abdominal thickness was measured in a transversal projection 1 cm above the umbilicus as the interspace between the internal surface of the rectus abdominis muscle and the anterior wall of the abdominal aorta (intra- and interobserver coefficients of variation, 4.1% and 4.8%, respectively). The maximal layer of preperitoneal fat (PreFT) was assessed below the xiphoid process in a transverse view. Finally, the ultrasound fat measurements were indexed by body weight through dividing the mean values by the BMI.

Statistical analysis

The distribution of variables was checked with the Kolmogorov–Smirnov test. Baseline clinical parameters and measures were compared between the groups using the t test for the normally distributed continuous variables and the Mann–Whitney test for the variables that did not follow normal distribution. Associations of individual parameters with the Homeostatic Model Assessment of Insulin Resistance (HOMA‑IR) and HbA_1c were assessed using the Spearman correlation analysis and were shown as R correlation coefficients. All demographic data, baseline characteristics, anthropometric parameters, vascular ultrasound indexes, and fat depots were analyzed as potential predictors of de novo DM using univariable logistic regression models. For each variable, receiver operating characteristic analysis was performed with the area under the curve estimation and selection of the cutoff value for the best binary prediction that maximized both sensitivity and specificity. Of all these variables, the ones with clinical importance or those indicating a possible impact were included in a multivariable logistic regression model. The final model was obtained using forward / backward stepwise regression with minimization of the Bayesian Information Criterion as the target. The final model vas validated using bootstrap resampling with 1000 iterations; then, a bias‑corrected C statistic was calculated as a measure of goodness‑of‑fit. A P value below 0.05 was considered significant. Statistical analysis was performed using MedCalc software (version 19.1, Ostend, Belgium).

Clinical characteristics of the study group

Detailed baseline (2012–2013) clinical characteristics of the study group are presented in Table 1. The participants comprised a high to very‑high cardiovascular risk population with a mean (SD) of 3.7 (1.0) risk factors, primarily arterial hypertension (85%), dyslipidemia (100%), and smoking (73%). Over one‑third of the study group were overweight (36.7%), and almost half (48.4%) were obese (obesity class I: 28.9%; class II: 12.5%; class III: 7%, respectively). Increased WC based on both the WHO and IDF definitions was found in 74 (59%) and 100 (80%) study participants, respectively. Detailed characteristics of anthropometric and ultrasound measures are presented in Table 2.

On baseline evaluation (2012–2013), DM was found in 28 patients (22%), and prediabetes (preDM) in 10 individuals (8%) (Table 1). All medications used initially in the study population are showed in Table 1.

Study follow‑up

After 8 years of follow‑up, death from any cause occurred in 15 patients (11.7%). Additionally, the number of DM cases increased to 52 (46%), and preDM was found only in 4 patients (3.5%). Moreover, 42 patients (37%) were treated with oral antidiabetic medications, and 18 (16%) used insulin.

Insulin resistance

The median HOMA‑IR was 3.02 (interquartile range [IQR], 2.02–4.98). There were 94 patients (73%) with a HOMA‑IR greater than or equal to 2.0 and 59 patients (46%) with a HOMA‑IR greater than or equal to 3.0. Among the ultrasound indexes, carotid EMT, IMT, and abdominal PreFT were found to significantly correlate with the HOMA‑IR. With respect to the anthropometric measures, BMI, WC, TCirc, total skinfold thickness, and body fat mass significantly correlated with the HOMA‑IR (Table 3). In the univariable analysis of ultrasound measures, a significant association between EMT and PreFT was found. Moreover, BMI, WC, waist‑to‑hip ratio (WHR), TCirc, and total skinfold thickness were also significantly linked with IR development (Table 4). Subsequently, in the multivariable analysis, only PreFT (per 10 mm increase; odds ratio [OR], 1.63; 95% CI, 1.22–2.33; P = 0.003) and a higher body surface area (per 0.1 m² increase; OR, 1.59; 95% CI, 1.11–2.39; P = 0.02) were independently associated with IR occurrence (Table 4). The positive (PPV) and negative (NPV) predictive values for the model were 0.71 and 0.84, respectively. The bias‑corrected C statistic was 0.87.

Type 2 diabetes

The median HbA_1c concentration was 5.9% (IQR, 5.6%–6.7%). Among the ultrasound indexes, a significant association between EMT or PreFT and HbA_1c was found. Additionally, among the anthropometric measures, BMI, TCirc, and body fat mass significantly correlated with HbA_1c levels (Table 3). In the univariable analysis, significant associations were demonstrated for EMT, IMT, PreFT, HDL‑C, BMI, WC, WHR, TCirc, total skinfold thickness, and body fat mass (Table 5). However, in the multivariable model, the correlation was only confirmed for HDL‑C (OR, 0.93; 95% CI, 0.87–0.97; P = 0.005) and body fat mass (OR, 1.09; 95% CI, 1.03–1.17; P = 0.003) (Table 5). The PPV and NPV for the model were 0.9 and 0.86, respectively, whereas the bias‑corrected C statistic was 0.80.

Multivariable regression analysis adjusted for clinical confounders

Finally, in a multivariable regression analysis of both anthropometric parameters and ultrasound indexes adjusted for age, estimated glomerular filtration rate, TC, LDL‑C, HDL‑C, and TG, we demonstrated that WC, WHR, total skinfold thickness, body fat, and body fat mass as well as EMT and PreFT were positively associated with the HOMA‑IR (Table 6). In turn, only WHR and body fat mass predicted an unfavorable increase in HbA_1c (Table 6).

Conclusions

A complex assessment of the adipose tissue in patients with CCS is a valuable method for improving cardiovascular risk stratification. As demonstrated, some anthropometric and ultrasound parameters can be helpful in the prediction of IR and DM development. Therefore, they could be incorporated into cardiovascular risk assessment algorithms.