MULTIMODAL LOGIT MODEL FOR PREDICTING THE EFFICIENCY OF MYOCARDIAL REVASCULARIZATION BY THE METHOD OF CORONARY ARTERY BYPASS GRAFTING IN PATIENTS WITH CORONARY HEART DISEASE.

Abstract

building of a mathematical logit model for possible prediction of the outcome of surgical treatment bythe method of coronary artery bypass grafting (CABG) in patients of different groups with coronary heart disease(CHD) based on myocardial viability (MV) assessment. To implement the set clinical tasks, 62 patients with coronary heart disease with preservedsystolic function and systolic dysfunction were examined. The mean age of the subjects was (59.6 ± 8.2) years. 35(56 %) patients had a variant of heart failure (HF) with an ejection fraction (EF) of 45 % or less. 27 (44 %) patientshad EF of 46 % or more. 5 (8.0 %) patients denied myocardial infarction (MI). Myocardial scintigraphy (MSG) wasperformed on Infinia Hawkeye combined gamma-camera (GE, USA) with integrated CT. The studies were performedin SPECT and SPECT / CT with ECG synchronization (Gated SPECT) modes. 99mTc-MIBI with an activity of 555-740 MBqwas used. MSG was performed in the dynamics of treatment (before CABG and after CABG) according to One Day Restprotocol. A total of 124 scintigraphic studies were performed. Samples of patients studied «before» and «after» the treatment were compared using nonparametricWilcoxon test (Wilcoxon Matched Pairs Test). A multivariate regression model, that reflects a statistically significanteffect on the treatment response (MV after treatment) of such cardiac activity indicators as LV EF (%), coronary bedlesion area and MV level (%) before treatment, was built. The above-described regression relationship between thethree above-defined functional factors of cardiac activity before treatment and the therapeutic effect in the formof the change in MV can be construed as a diagnostic model that predicts the treatment outcome. This scientific study allows to build logit models to predict the expected outcome of coronary heartdisease surgical treatment in patients of different groups. The presented multivariate regression model is characterised by a sufficiently high for biostatistical studies adjusted coefficient of determination (Adjusted R2 = 0,893 (F = 173,4; p < 0,001)).