Establishing a Non-Invasive Gradeing Model for Primary IgA Nephropathy Based on Multi-Center Population.

Abstract

It is necessary to adopt a special therapeutic schedule for treating IgA nephropathy (IgAN) at a distinct pathological stage. It would be helpful for treatment to know the pathological changes throughout the IgAN course; therefore, we want to establish a non-invasive method for determining the pathological grade of IgAN. A total of 240 primary IgAN patients were recruited, and their clinical data and laboratory test results were collected. The study subjects were randomly divided into the training set (181 cases) and testing set (59 cases). The ordered logistic regression model was constructed with variables which were selected by single-factor and multi-factor stepwise regression analysis in training set, then the model was verified by testing set. Logistic regression analysis showed that hematuria, hemoglobin, urea, complement 3, urinary microalbumin and urinary microalbumin/creatinine are related to Lee's classification of IgAN. Using the above indicators as independent variables to establish the non-invasive grading model. The model's accuracy is as high as 82.9% (p = 0.00), and the rate of precision, recall, and specificity for each group are all above 80%. This model discriminates four classes of pathological stage corresponding to Lee's grading well. A non-invasive grading model for primary IgAN has been established successfully by clinical and laboratory data.