Urinary sediment microRNAs can be used as potential noninvasive biomarkers for diagnosis, reflecting the severity and prognosis of diabetic nephropathy

Abstract

Abstract Background Patients with both diabetes mellitus (DM) and kidney disease could have diabetic nephropathy (DN) or non-diabetic renal disease (NDRD). IgA nephropathy (IgAN) and membranous nephropathy (MN) are the major types of NDRD. No ideal noninvasive diagnostic model exists for differentiating them. Our study sought to construct diagnostic models for these diseases and to identify noninvasive biomarkers that can reflect the severity and prognosis of DN. Methods The diagnostic models were constructed using logistic regression analysis and were validated in an external cohort by receiver operating characteristic curve analysis method. The associations between these microRNAs and disease severity and prognosis were explored using Pearson correlation analysis, Cox regression, Kaplan–Meier survival curves, and log-rank tests. Results Our diagnostic models showed that miR-95-3p, miR-185-5p, miR-1246, and miR-631 could serve as simple and noninvasive tools to distinguish patients with DM, DN, DM with IgAN, and DM with MN. The areas under the curve of the diagnostic models for the four diseases were 0.995, 0.863, 0.859, and 0.792, respectively. The miR-95-3p level was positively correlated with the estimated glomerular filtration rate (p < 0.001) but was negatively correlated with serum creatinine (p < 0.01), classes of glomerular lesions (p < 0.05), and scores of interstitial and vascular lesions (p < 0.05). However, the miR-631 level was positively correlated with proteinuria (p < 0.001). A low miR-95-3p level and a high miR-631 level increased the risk of progression to end-stage renal disease (p = 0.002, p = 0.011). Conclusions These four microRNAs could be noninvasive tools for distinguishing patients with DN and NDRD. The levels of miR-95-3p and miR-631 could reflect the severity and prognosis of DN.