Gender differences in fasting and postprandial metabolic traits predictive of subclinical atherosclerosis in an asymptomatic Chinese population
The study took place at two sites: the Singapore National Heart Center (NHCS) for cardiovascular measurements, and the Singapore Institute of Clinical Sciences (SICS), A*STAR for the mixed meal test. Noninvasive imaging was used to assess the presence of AS in 101 healthy Chinese adults recruited between May 2018 and June 2019. Inclusion criteria were: (1) be willing and able to sign informed consent written in English or Chinese before trial entry; (2) aged 40 to 54; (3) male or female; (4) Chinese ethnic group (having both Chinese grandparents); (5) low Framingham risk of CAD (2); (7) previous myocardial infarction; (8) known prior coronary revascularization; (9) known and documented peripheral arterial disease; (10) previous stroke (defined as a new focal neurological deficit persisting > 24 h); (11) use of antihypertensive agents; (12) history of cancer (excludes precancerous lesions); (13) life expectancy fasting and postprandial blood phenotyping. The study protocol was approved by the SingHealth Centralized Institutional Review Board (2018/2116) and all eligible participants provided written informed consent. This trial has been registered with ClinicalTrials.gov (NCT03531879). All procedures were conducted ethically in accordance with the Declaration of Helsinki.
Definition of a low FRS
Subjects in this study were at low risk (≤10%) of developing coronary artery disease as defined by the Framingham Heart Study 10-year risk score.23. Fasting blood pressure, smoking status, total cholesterol, low-density lipoprotein cholesterol (LDL-C), and high-density lipoprotein cholesterol (HDL-C) were used to calculate the risk score at 10 years of coronary heart disease.
The presence of atherosclerotic plaque in bilateral carotid arteries, bilateral ilio-femoral arteries and abdominal aorta was determined by duplex ultrasound (Philips EPIQ-7G ultrasound system and iU22 ultrasound system 795050). Plaques were defined as a focal protrusion into the arterial lumen with a thickness >0.5 mm or >50% of the surrounding intima-media thickness or a diffuse thickness >1.5 mm measured between the media-adventitia interfaces and intima-light24. Coronary artery calcification (CAC) was measured by non-contrast chest CT to calculate the Agatston score (Canon Scanner Aquilion-One, Genesis-Edition, Model: TSX-305A/55)25. AS was defined as the presence of atherosclerotic plaque in the carotid, abdominal aortic, or iliofemoral territory and/or a CAC Agatston score ≥ 1. Participants with 0 affected vascular sites were classified as free (plaque = 0). Participants with at least 1 affected plaque site (plaque >0) were considered to have AS.
Within 2 weeks of the cardiovascular imaging assessment, subjects attended SICS for the mixed meal test. Blood biomarkers were measured after 10 h of fasting and at frequent intervals for 6 h after consumption of the test meal. Venous blood samples were collected in precooled sterile tubes (Vacutainer, Becton Dickinson) containing K2EDTA (final concentration 1.8 mg/mL) and a cocktail of protease inhibitors (AG Scientific, Inc, CA, USA) for plasma samples, and sterile tubes containing silica clot activator for serum samples. Plasma was collected by low-speed centrifugation (1750×g10 mins, 4°C).
Venous blood was collected on an empty stomach and at 10, 20, 30, 45, 60, 90, 120, 240 and 360 min after consumption of the test meal for the determination of insulin, glucose and C-peptide, in due to the acute insulin response in which more than 90% of the pancreatic beta cell response occurs within the first 10 to 30 minutes. Blood samples were taken on an empty stomach and at 60, 120, 240 and 360 min for the determination of apolipoprotein B48, leptin, adiponectin, C-reactive protein (CRP), tumor necrosis alpha (TNFα), interleukin 6 (IL-6), plasminogen activator inhibitor-1 (PAI-1), vascular cell adhesion molecule 1 (VCAM-1), adhesion molecule intercellular 1 (ICAM-1), E-selectin, LDL-C, HDL-C, total cholesterol and triacylglycerol. Plasma glucose concentration was determined using the glucose oxidase method on an automated glucose analyzer (YSI 2300 Stat Plus; YSI Life Sciences, Yellow Spring, OH, USA). Plasma insulin and C-peptide concentrations were determined by electrochemiluminescence technology (Roche/Hitachi cobas e411 immunochemistry analyzer; Roche Diagnostics, Indianapolis, IN). Apolipoprotein B48, leptin, adiponectin, CRP, TNFα, IL-6, PAI-1, VCAM-1, ICAM-1 and E-selectin were determined using a 6-plex Human Luminex Screening Human Magnetic Bead Immunoassay (R&D Systems, Inc.), a 3-Plex Luminex Screening Human Magnetic Assay (R&D Systems, Inc.) and an ELISA kit for human apolipoprotein B48 ( Elabscience Biotechnology Inc.). Data analysis was performed on Bio-Plex Manager™ 6.1.1 (Bio-Rad). Standard curves were generated with a 5-parameter logistic algorithm, reporting values for mean fluorescence intensity and concentration data.
The test meal consisted of 237.5 ml of a nutritional drink (Nestlé Heath Sciences, Lausanne, Switzerland) mixed with 100 ml of commercially available whipping cream. This composition was determined based on a systematic review that defined an optimal nutritional stress test26: 75 g of glucose, 60 g of palm olein and 20 g of dairy protein served in a liquid meal of ~ 337 mL providing a total of ~ 930 kcal.
Statistical comparisons between participants with and without AS were made using Fisher’s exact test for categorical variables and Student’s t-test. you comparison test of two independent samples in the evaluation of continuous variables. Univariate and multivariate logistic regression models incorporating a stepwise variable selection approach for model building were used to analyze the associations of several covariates with the presence of AS and to select predictors.
To estimate the biomarker response during the entire 6 h postprandial period, the incremental AUC over the fasting level was calculated. Associations between metabolic markers and the presence of atherosclerosis were then assessed using univariate logistic regression. Variables that showed marginal univariate association with pPp= 0.10 to enter and stay. ROC curves were obtained from multivariate logistic regression models, and Youden’s index was used to find the statistically optimal threshold to predict high or low risk. The statistically optimal threshold for men was a predicted probability of P= 0.39 corresponding to the Youden index of J = 0.727; for females, P= 0.55, corresponding to J = 0.832; and for pooled samples, P= 0.35 and J = 0.543. The predictive models with corresponding ROC curve summaries, AUCs, predictive statistics, and optimal threshold discrimination abilities are summarized in the figures and tables below. The predictive models were then subjected to tenfold cross-validation to obtain a deoptimized ROC curve. We used the Delong-Delong approach to compare the ROC AUC between the Framingham score model and the postprandial biomarker predictive model. Cross-validation was performed by calling the CROSSVALIDATE option of the SAS logistic regression procedure (PROC LOGISTIC) which requests the validated individual predicted probability of each level of response. These probabilities are derived from the opt-out principle, that is, by deleting a subject’s data and re-estimating the parameter estimates. However, PROC LOGISTIC uses a computationally less expensive one-step approximation to calculate parameter estimates that are valid for binary response models. All statistical analyzes were performed using SAS University Edition (SAS Institute Inc., NC, USA). Unless otherwise stated, statistical significance was set at p