CARDIOMETABOLIC RISK PREDICTION IN PATIENTS WITH NON-ALCOHOLIC FATTY LIVER DISEASE COMBINED WITH SUBCLINICAL HYPOTHYROIDISM

. One of the most common diseases of our time is non-alcoholic fatty liver disease (NAFLD). Recently published research results indicate that patients with NAFLD along with traditional risk factors for cardiovascular diseases (CVD) have "new" risk factors such as endothelial dysfunction (ED), carotid intima-media thickness (CIMT), an increase in the CRP level, as well as risk factors combined into the Framingham scale. It is also known that combination of NAFLD with subclinical hypothyroidism (SH) forms an abnormal metabolic phenotype, which is associated with cardiometabolic risk factors. The study of cardiometabolic predictors and vascular markers in patients with NAFLD in combination with SH will provide an opportunity to improve the strategy of cardiovascular events prevention in such comorbid patients.


Introduction
In recent years, an increasing number of cardiovascular pathologies worldwide makes it necessary to search for new risk factors that could serve as an indicator and determine the likely prognosis of patients with somatic diseases. Currently, the concept of "cardiometabolic risk" is increasingly used in clinical practice. It involves the risk of developing cardiovascular diseases and/or type 2 diabetes, combining both classical ("traditional") risk factors such as smoking, high levels of low-density lipoprotein cholesterol (LDL-C), arterial hypertension, an increase in glucose level, and factors directly associated with abdominal (especially visceral) obesityinsulin resistance (IR), low high-density lipoprotein cholesterol (HDL-C), hypertriglyceridemia, and an increase in proinflammatory markers [8]. The attention of researchers is attracted by the search for early predictors of cardiometabolic changes and vascular markersmarkers of inflammation (C-reactive protein, CRP), dyslipidemia, IR, tumor necrosis factor -α (TNF-α), endothelial dysfunction (ED) indicators such as circulating desquamated endothelial cells (CDECs), vascular endothelial growth factor (VEGF), that could act as a "predictive" [1,7].
One of the most common diseases of our time is non-alcoholic fatty liver disease (NAFLD)a chronic liver disease that is considered as a component of some diseases associated with IR, such as metabolic syndrome (MS), obesity, and diabetes. The development of NAFLD is very closely associated not only with abdominal obesity, but also has a significant effect on the formation of cardiometabolic risk factors such as hypertriglyceridemia, a decrease in HDL-C level, hypertension, hyperglycemia, thereby increasing the degree of cardiometabolic risk itself and affecting the prognosis and life expectancy of patients [8,18].
Recently published research results indicate that patients with NAFLD along with traditional risk factors for cardiovascular diseases (CVD) (obesity, diabetes, MS and others) have "new" risk factors such as endothelial dysfunction (ED), carotid intimamedia thickness (CIMT), an increase in the CRP level, as well as risk factors combined into the Framingham scale (age, gender, hypertension, hyperlipidemia, smoking). Also, "traditional" cardiovascular risk factors include: markers of chronic inflammation, hyperhomocysteinemia, impaired blood viscosity, as well as substances that the endothelium secretes (NO, Willebrand factor, P-selectin, adhesion molecule ICAM-1 and VCAM-1, endothelin-1, PAI-1) [7,8].
Atherogenic dyslipidemia already occurs during SH and the early stages of NAFLD. It is characterized by an increase in the level of very low-density lipoprotein cholesterol (VLDL-C), LDL-C, and may also occur due to the thyroid hormone deficiency, which leads to a decrease in LDL receptors in the liver, as a result of a decrease in hepatic total cholesterol excretion, and then -to an increase in LDL-C and VLDL-C levels. The combination of NAFLD with SH forms an abnormal metabolic phenotype, which is characterized by the presence of dyslipidemia, hyperinsulinemia and IR associated with oxidative stress and endothelial dysfunction [12,14].
Thus, the study of cardiometabolic predictors and vascular markers in patients with NAFLD in combination with SH will provide an opportunity to broaden understanding of the mechanisms of cardiometabolic risk formation and to individualize the strategy of cardiovascular events prevention in a comorbid patient within this framework.
One of the rapidly progressing trends in clinical medicine is undoubtedly the predicting of diseases using special scales. The SCORE scale (or the electronic version of HeartScore), popular in Europe, is limited to the age of 45-64 years, while the age range in the PCE scale was much wider -20-79 years old, but it is recommended to use it in the range of 40-79 years [3,19].
Solving the problems of medical forecasting often involves the use of regression models. For example, Weibull regression was used in the SCORE scale, Cox regression was used in the second version of the GRACE scale, and the logistic regression was used in the CHA2DS2-VASc scale [2,20]. One of the principal limitations of the scales is the impossibility of an individual assessment, since scales can only give a probabilistic risk assessment for a group of patients with given levels of risk factors.
Preventive treatment of CVD is largely based on model creation methods for assessing the "absolute risk" of future cardiovascular events in order to substantiate decisions about therapeutic approaches [5].
Neural networks widely used in medical science. Kojuri J. et al. [6] reported on the development of a neural network that predicts acute myocardial infarction for two weeks.
Jasnickij L. and Cherepanov F. [17] showed in their article the possibility of creating neuroexpert medical systems, allowing to perform long-term prognosis of the diseases development to predict the occurrence of new diseases in future periods of the patient's life.
The introduction of machine learning allows increasing the accuracy of prediction of cardiovascular events [15]. However, the use of a large array of heterogeneous data instead of selected patient cohort can reduce the accuracy of risk classification and the proximity of the forecast to the frequency of real events [4,9,13,16].
The subject of the present work is creating a model for predicting cardiometabolic risk in patients with NAFLD in combination with SH.
The paper presents the results of the GI "L.T. Malaya Therapy Institute of the National Academy of Medical Science of Ukraine" scientific research on the topic: "Influence of factors of cardiovascular development on premature aging" (security number 0117U003031).

Materials and methods
71 patients were selected to create a prediction model for cardiometabolic risk in patients with NAFLD in combination with SH. All patients were divided into three groups as follows: Group 1low risk patients (6 patients); Group 2patients with moderate risk (43 patients); Group 3high-risk patients (22 patients).
Mathematical processing of the results was carried out using the SPSS19 application package for Windows.

Results of binary regression logistics analysis
49 patients of the first and second groups were analyzed to construct an equation of logistic regression, which determines the probability of having a moderate cardiovascular risk (CVR), taking into account the considered indicators.
Comparison of groups 1 and 2 showed that among all the evaluated factors, the length of the telomeres in the buccal epithelium, the levels of CRP, CDECs had the statistically significant effect on the probability of development of moderate cardiometabolic risk in patients with NAFLD in combination with SH. As a result, the regression function was constructed. It included 3 indicators (  (1) where ̂ is the probability of developing moderate CVR in patients with NAFLD in combination with SH. The calculated coefficients of the regression function and the results of checking their significance are presented in table 2. All variables are significant (p < 0.05) and selected correctly according to Wald's statistics. The quality of the regression model approximation was estimated using the similarity function. In our study, G = 11,612 at p = 0.001, that is as a whole the informative indications isolated as independent variables have a significant effect on the development of cardiometabolic risk in patients with NAFLD in combination with SH. The Cox and Schell, and Nagelkerke indices indicate that the dispersion part explained by the developed logistic model stands at 75.8% (table 3). The figure 1 shows the classification diagram. The distribution point was P = 0.5. The closer the value of the predicted probability is to unit, the higher the probability of developing moderate cardiometabolic risk in patients with NAFLD in combination with SH is. The results of the classification of patients in a training group using the developed prognostic model are given in table 4. 5 patients of the total number of patients with low risk, equal to 6, were correctly and 1 was mistakenly attributed to the moderate risk group.
Of the total number of patients with moderate risk of 43 people, 42 patients were accurately identified and 1 was IAPGOŚ 2/2023 p-ISSN 2083-0157, e-ISSN 2391-6761 erroneously assigned to a group of low-risk patients. In general 47 cases out of 49 were correctly identified, representing 95.9%.
A general assessment of the agreement between the influence of risk factors found in the model and the actual occurrence of an adverse outcome was carried out using the Hosmer-Lemeshow (HL) test.
The resulting low value of HL = 4.391 at a significance level of p > 0.05 (p = 0.820), indicates the minimum differences between observed and predicted frequencies and the high quality of the selected regression model. The ROC analysis of the obtained model (figure 2) confirmed its effectiveness and excellent predictive quality. The area under curve (AUC) was 0.977 at p = 0.001.

Fig. 2. ROC Curve
Thus, a mathematical model has been developed that includes metabolic risk factors such as the length of telomeres in the buccal epithelium, CRP and CDECs, to determine effectively and qualitatively the probability of forming moderate cardiometabolic risk in patients with NAFLD in combination with SH.
Data from 65 patients in the second and third groups were analyzed for the purpose of constructing a prediction model of cardiometabolic risk of higher gradations in patients with NAFLD in combination with SH. As a result of the binary logistic regression analysis, an equation has been obtained, which determines the likelihood of developing high CVR in patients with NAFLD in combination with SH:  (2) where ̂ is the probability of developing high CVR in patients with NAFLD in combination with SH; Х 1telomere length in the blood; Comparison of 2nd and 3rd groups therefore showed that of all of the factors evaluated, telomere length in blood, LDL-C, VEGF and TSH had a statistically significant effect on the likelihood of high cardiometabolic risk developing in patients with NAFLD in combination with SH.
All variables are significant (p < 0.05) and selected correctly according to Wald's statistics (table 5). In the study, G = 39.297 at p = 0.001 shows that independent variables have a significant contribution to predict the dependent variable in general. The part of the dispersion, explained by logistic regression according to R² of Cox and Schell stands at 68% (table 6). A chart of classification of patients in the second and third study groups is presented in fig. 3 with the help of the developed mathematical model for predicting high cardiometabolic risk in patients with NAFLD in combination with SH. The value of P = 0.5 served as the limiting point of distribution. The closer the value of the predicted probability is to unity, the higher the degree of cardiometabolic risk in patients with NAFLD in combination with SH is. The classification results presented in table 7 indicate that in general 59 cases were correctly identified out of 65 that is 90.8%. The  Thus, blood serum indices -LDL-C; VEGF; TSH and the length of telomeres in the blood may be factors that determine the probability of forming cardiometabolic risk of higher gradations in patients with NAFLD and SH.

Validative review of developed prognostic models
The usefulness of the developed mathematical models has been tested in practice while predicting the development of cardiometabolic risk in patients with NAFLD in combination with SH, who were screened and treated at the Government Institution "L.T. Malaya Therapy Institute of the National Academy of Medical Science of Ukraine". 34 patients were examined and divided into the following groups: Group 1moderate-risk patients (9 patients); Group 2high-risk patients (17 patients). Group 3high-risk patients (8 patients).
The results of testing the model of binary logistic regression, created to determine the probability of moderate CVR development in patients with NAFLD in combination with SH, are presented in table 8. According to the results obtained, the outlook for 25 examined patients was made unmistakably. Thus, it was confirmed in 96.2% of cases.
The results of checking the model of binary logistic regression, created to determine the likelihood of high CVD in patients with NAFLD in combination with SH, are presented in table 9. The outlook was unmistakably accurate for 24 patients surveyed according to the results. Thus, the forecast was confirmed in 92.0% of cases.
That means, that the application of synthesized logistic models will allow predicting the development of cardiometabolic risk, which will ensure early diagnosis and appointment of treatment and prevention measures in order to avoid the development of cardiometabolic complications of high gradations.

Conclusions
The developed mathematical models allow predicting the degree of cardiometabolic risk in patients with NAFLD in combination with SH based on the of cumulative changes in lipid, carbohydrate states, thyroid function compensation, vascular factors and telomere lengths as a marker for vascular aging. The application of the proposed prognostic models in clinical practice allows to achieve an improvement in the quality of the cardiometabolic risk determination in patients with NAFLD in combination with SH due to the identification of factors that influence the progression of cardiometabolic changes in patients with NAFLD precisely in combination with SH, and the improvement of the applied mathematical apparatus, which in turn will help the physician to prescribe adequate treatment and preventive measures and improve the quality of medical services provision for patients with NAFLD in combination with SH.