Predictive potential of cardiovascular risk factors and their associations with arterial stiffness in people of European and Korean ethnic groups

Abstract

Aim. To compare the effect of cardiovascular risk factors on aortic stiffness in people of European and East Asian ethnic groups.Material and methods. A total of 266 patients aged 18-60 years of European (n=133) and Korean (n=133) ethnic groups were examined. Clinical assessment was carried, Also, following blood parameters was evaluated: total cholesterol (TC), low- (LDL-C) and high- (HDL-C) density lipoprotein cholesterol, apolipoproteins A (apo-A) and B (apo-B), triglycerides (TG), uric acid, creatinine, glucose, adiponectin, resistin. The aortic pulse wave velocity (PWV) and central blood pressure (CBP) were determined using a Tensiomed arteriograph (Hungary). The study design included 3 stages. The first stage included statistical analysis using Mann-Whitney, χ2, Fisher tests, while the second one — determination of weighing coefficients of individual risk factors on aortic PWV. The third stage consists of verification of the relationship between ethnicity and aortic PWV using multivariate logistic regression and stochastic gradient boosting (SGB).Results. In Europeans, the median values of growth, body mass index (BMI), waist circumference (WC) and waist-to-height ratio were significantly higher, while the levels of apo-B, TC, HDL-C, LDL-C, TG was significantly lower than in Asians. Koreans had higher blood concentrations of UA, creatinine, glucose, while the resistin concentration was 1,8 times lower. Among Europeans, the odds ratio of developing hypertension (HTN) was significantly higher. The level of aortic PWV in people of different ethnic groups did not differ significantly. Univariate logistic regression showed a dominant influence of age, CPP and waist-to-height ratio on aortic PWV. A less noticeable significant relationship with aortic PWV had HTN, female sex, BMI, levels of systolic, diastolic and pulse BP. Multivariate logistic regression and SGB showed the maximum prediction accuracy when 5 predictors were combined in one model: age, height, HTN, LDL-C, and ethnicity. Comparable accuracy was demonstrated by a model where glucose level was used instead of LDL-C. The results indicate a nonlinear relationship between the ethnic factor and aortic PWV. Its predictive potential was realized only in combination with functional and metabolic status parameters of patients. In Koreans, the threshold values of these factors can be significantly higher than in Europeans.Conclusion. Developed using modern machine learning technologies, the assessment aortic PWV models taking into account the ethnic factor can be a useful tool for processing and analyzing data in predictive studies.

Keywords

• cardiovascular risk
• ethnicity
• aortic stiffness
• machine learning
• mathematical models