Validation of a differential diagnostic model of diabetic nephropathy and non-diabetic renal diseases and the establishment of a new diagnostic model

Background: The aims of the present study were to validate the differential diagnostic model of diabetic nephropathy (DN) and non-diabetic renal diseases (NDRD) established in 2003 and to establish a new diagnostic model suitable for the current clinical characteristics of DN. Methods: We examined 200 patients with Type 2 diabetes who underwent kidney biopsy from 2004 to 2012. The 2003 differential diagnostic model based on the data collected from 1993 to 2003 was evaluated by the diagnostic test and changes in the clinical differentiation parameters of DN and NDRD were analyzed. Logistic regression, receiver operating characteristics (ROC) curve, net reclassification improvement (NRI), and integrated discrimination improvement (IDI) analysis were applied. Results: The 2003 diagnostic model showed an accuracy of 77.5%. A significantly elevated incidence of hematuria, longer history of diabetes, and reduced level of glycated hemoglobin (HbA1c) were observed in the DN group from 2004 to 2012 compared with DN group from 1993 to 2003. Histories of diabetes mellitus (Dm), systolic blood pressure (Bp), HbA1c (Gh), hematuria (Hu), diabetic retinopathy (Dr), and hemoglobin (Hb) are independently related to DN. Thus, a new diagnostic model was constructed as follows: P-DN = exp (0.846 + 0.022 Dm + 0.033Bp + 2.050 Gh-2.664 Hu-0.078 Hb + 2.942Dr)/[1 + exp (0.846 + 0.022 Dm + 0.033 Bp + 2.050 Gh-2.664 Hu-0.078 Hb + 2.942 Dr)]. Validation tests determined that the accuracy of the new model were 90.9%. Conclusions: Changes in people with DN, clinical characteristics have reduced the diagnostic efficacy of the 2003 diagnostic model. The newly established model can provide a better, more current differentiation between DN and NDRD.