Application of DataMining clustering and classification methods in the study of the mechanisms of mutual burdening of reactive arthritis and chronic pyelonephritis in terms of connective tissue metabolism

Abstract

Disorders of metabolic processes in the "proteoglycans-collagen" system, changes in enzymatic reactions in patients with reactive arthritis (ReA) are often preceded by many complications, including impaired renal function. Classical regression methods are not very informative in determining the mutual burden of ReA and chronic pyelonephritis (CP), as well as the primacy of the disease.Objective - to investigate the possible relationship between ReA and CP in order to establish early criteria for predicting the development of CP on the background of ReA on the indicators of connective tissue metabolism. Material and methods. 113 patients were examined, which were divided into two groups: the first group - patients with urogenic ReA, activity I-III. FTS I-III st. (n = 65); the second group - patients with urogenic ReA and CKD I-II: pyelonephritis in the acute phase (n = 48). The control group consisted of 20 healthy individuals. The average duration of the disease of the examined patients was 24.4 ± 4.7 months. The mean age of patients was 32.5 ± 1.2 years. DataMining clustering and classification methods were used to process the obtained research results.Results. As a result of clustering methods (k-means and fuzzy clustering), the same results of cluster membership were obtained. In particular, ReA disease was correctly diagnosed in 48 cases, which is 74% of all diagnosed, 17 people (26%) were assigned to cluster "2". There is also a mismatch in the second cluster. In particular, out of 32 people diagnosed with ReA + CP, according to mathematical calculations, 28 people (88%) got into this cluster. Four patients (13%) were assigned to cluster "1" – to the group of persons in whom only one ReA disease should be diagnosed. This indicates that the boundary between the ReA and ReA + CP clusters are somewhat blurred. And this is the basis for establishing the fact that ReA disease can gradually lead to CP.Conclusions. Using DataMining clustering methods the greatest significance was defined in the diagnostic algorithm of ReA progression of such indicators as free oxyproline (FOP), protein-bound oxyproline (PBOP), the degree of collagenolytic activity (CLA) (intensity of azocol lysis), which showed a direct dependence on the degree of activity of the inflammatory process. Increases in FOP> 13.8 μmol/l, PBOP> 65.0 μmol/l and CLA (azocol)> 0.85 μg / ml in 1 h are probable risk factors for progression and early criteria for severe ReA and CP.