ФЕНОТИПОВІ ОЗНАКИ І ПОЛІМОРФІЗМ RS10465885 ГЕНА КОНЕКСИНУ-40 ЯК ПРЕДИКТОРИ ВИНИКНЕННЯ РЕЦИДИВУ АРИТМІЇ У ПАЦІЄНТІВ З ФІБРИЛЯЦІЄЮ ПЕРЕДСЕРДЬ НЕКЛАПАННОГО ҐЕНЕЗУ ПІСЛЯ ВІДНОВЛЕННЯ СИНУСОВОГО РИТМУ ПРИ 1-РІЧНОМУ СПОСТЕРЕЖЕННІ: РЕЗУЛЬТАТИ НЕЙРОМЕРЕЖЕВОГО АНАЛІЗУ

Atrial fibrillation (AF) is a violation of the heart rhythm associated with 1 % (or even more) of the health budget expenditure in Europe. Despite the fact that there is a constant increase in the socio-economic burden of AF, and there has been significant progress in understanding the pathophysiology of this arrhythmia, the effectiveness of its treatment is still far from satisfactory. One of the reasons for the insufficient effectiveness of modern strategies for the prevention and treatment of AF is the existence of limitations in understanding the complex pathophysiology of this disturbance of the heart rhythm.The aim of the study – to leran the phenotypic and genotypic (rs10465885 polymorphism in connexin-40 [Cx40] gene) predictors of arrhythmia recurrence in patients with non-valvular atrial fibrillation (AF) after sinus rhythm restoring (SRR ) at 1-year follow-up (AF360), in particular by the use of artificial neural networks (ANN ) analysis.Materials and Methods. We enrolled 104 patients (pts) with non-sustained non-valvular AF (average age (53±10) years, 80 (76.9 %) men). The distribution of rs10465885 polymorphic variants in Cx40 gene (n=73) was as follows: TT – 17 (23.3 %) pts, CT – 33 (45.2 %), СС – 23 (31.5 %). We analyzed 122 cases of SR restoring: 32 (26.2 %) – pharmacological cardioversion (29 pts); 63 (51.6 %) – electrical cardioversion (53 pts); 27 (22.2 %) – radiofrequency catheter ablation (22 pts). AF360 occurred in 76 (65.5 %) of 116 cases. In order to identify AF360 predictors, we used logistic regression analysis, as well as ANN analysis (building the linear [LIN] and nonlinear [multilayer perceptron (MLP )] ANN`s).Results and Discussion. Genetic algorithm Input Selection revealed 16 parameters, associated with AF360, including SRR type, rs10465885 polymorphism, and certain clinical (in particular, CHA 2DS 2-VAS c score), lab and echo parameters. CHA 2DS 2-VAS c score was the only independent AF360 predictor, according to the results of multivariable logistic regression analysis. The area under curve (AU C) for MLP , included all 16 variables, revealed by the Genetic algorithm, was significantly higher than in linear LIN: 0.874 (95 % confidence interval [CI] 0.798–0.929) vs. 0.678 (CI 0.583–0.763), respectively (p<0.001). In order to obtain the maximal reduction of predictors, we built the MLP , included the set of 5 variables (MLP 5): SRR type; rs10465885 polymorphism; the heart failure presence and its stage; AF type (recurrent or first diagnosed); and antero-posterior left atrial dimension. MLP 5 AU C (0.808 (95 % CI 0.723–0.876)) was significantly higher than those for LIN (р = 0.027).Conclusion. AF360 was non-linearly associated with SRR type, rs10465885 polymorphism in Cx40 gene, as well as certain phenotypic parameters. The further search of the most significant genetic and epigenetic predictors of AF recurrence at different terms after SRR is of crucial importance.