Prediction of subclinical gouty nephropathy by using neural networks

Abstract

Gouty nephropathy may have a subclinical course for years, but with what probability may it occur? We used neural networks or “artificial intellect” to solve this problem. This study aims to recognize the risk factors that correlate with significantly high microprotein indices and build a neural network that will predict the development of subclinical course of gouty nephropathy in percentage without additional tests. A one-center cohort prospective study included 117 gouty arthritis patients who were on scheduled in-patient treatment at the rheumatology department during 2018–2021. All patients had no history of ongoing kidney disease. The study aimed to recognize the risk factors that correlate with significantly high microprotein indices and build a neural network that will predict the development of subclinical course of gouty nephropathy in percentage without additional tests. We can distinguish the factors that are most associated with the development of kidney disease and correlate with microalbumin (r1) and α1-microglobulin (r2) in the urine: hyperuricemia (r1=0.85; r2=0.73), hypouricosuria (r1=-0.79; r2=-0.63), hypertriglyceridemia (r1=0.84; r2=0.78), an increase in LDL levels (r1=0.77; r2=0.79). There were also established correlations between renal disease and the fact of arterial hypertension (r=0.81), diabetes mellitus (r=0.59). The neural network was built. That is why, having input data, it is easy to predict the risk of gouty nephropathy development by using the proposed calculator and beginning early target prophylaxis, even if kidney disease may have a subclinical course.