AUTHOR=Xu Lin-Lin , Zhang Di , Weng Hao-Yi , Wang Li-Zhong , Chen Ruo-Yan , Chen Gang , Shi Su-Fang , Liu Li-Jun , Zhong Xu-Hui , Hong Shen-Da , Duan Li-Xin , Lv Ji-Cheng , Zhou Xu-Jie , Zhang Hong 

TITLE=Machine learning in predicting T-score in the Oxford classification system of IgA nephropathy

JOURNAL=Frontiers in Immunology

VOLUME=14

YEAR=2023

URL=https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2023.1224631

DOI=10.3389/fimmu.2023.1224631

ISSN=1664-3224

ABSTRACT=<sec><title>Background</title><p>Immunoglobulin A nephropathy (IgAN) is one of the leading causes of end-stage kidney disease (ESKD). Many studies have shown the significance of pathological manifestations in predicting the outcome of patients with IgAN, especially <italic>T</italic>-score of Oxford classification. Evaluating prognosis may be hampered in patients without renal biopsy.</p></sec><sec><title>Methods</title><p>A baseline dataset of 690 patients with IgAN and an independent follow-up dataset of 1,168 patients were used as training and testing sets to develop the pathology <italic>T</italic>-score prediction (<italic>T</italic><sub>pre</sub>) model based on the stacking algorithm, respectively. The 5-year ESKD prediction models using clinical variables (base model), clinical variables and real pathological <italic>T</italic>-score (base model plus <italic>T</italic><sub>bio</sub>), and clinical variables and <italic>T</italic><sub>pre</sub> (base model plus <italic>T</italic><sub>pre</sub>) were developed separately in 1,168 patients with regular follow-up to evaluate whether <italic>T</italic><sub>pre</sub> could assist in predicting ESKD. In addition, an external validation set consisting of 355 patients was used to evaluate the performance of the 5-year ESKD prediction model using <italic>T</italic><sub>pre</sub>.</p></sec><sec><title>Results</title><p>The features selected by AUCRF for the <italic>T</italic><sub>pre</sub> model included age, systolic arterial pressure, diastolic arterial pressure, proteinuria, eGFR, serum IgA, and uric acid. The AUC of the <italic>T</italic><sub>pre</sub> was 0.82 (95% CI: 0.80–0.85) in an independent testing set. For the 5-year ESKD prediction model, the AUC of the base model was 0.86 (95% CI: 0.75–0.97). When the <italic>T</italic><sub>bio</sub> was added to the base model, there was an increase in AUC [from 0.86 (95% CI: 0.75–0.97) to 0.92 (95% CI: 0.85–0.98); <italic>P</italic> = 0.03]. There was no difference in AUC between the base model plus <italic>T</italic><sub>pre</sub> and the base model plus <italic>T</italic><sub>bio</sub> [0.90 (95% CI: 0.82–0.99) <italic>vs</italic>. 0.92 (95% CI: 0.85–0.98), <italic>P</italic> = 0.52]. The AUC of the 5-year ESKD prediction model using <italic>T</italic><sub>pre</sub> was 0.93 (95% CI: 0.87–0.99) in the external validation set.</p></sec><sec><title>Conclusion</title><p>A pathology <italic>T</italic>-score prediction (<italic>T</italic><sub>pre</sub>) model using routine clinical characteristics was constructed, which could predict the pathological severity and assist clinicians to predict the prognosis of IgAN patients lacking kidney pathology scores.</p></sec>