Non-Invasive Prediction Tool for Non-Diabetic Kidney Diseases in Patients with Type-2 Diabetes Mellitus

Background Despite being the gold standard in detecting non-diabetic kidney diseases (NDKD) in Type-2 Diabetes Mellitus (T2DM), renal biopsy poses an inherent risk of life-threatening complications. The current study aims to develop and validate a non-invasive scoring tool to predict NDKD using clinical and laboratory variables. Materials and Methods We developed a model to detect NDKD using multivariable binary logistic regression analysis with the backward Wald elimination method. We included all patients with T2DM who had an indication kidney biopsy for NDKD during the study. The model was assessed using the area under curve-reciever operating characteristic (AUC-ROC) curve on both the derivational and validation cohort and by multicentric external validation. Results Out of 538 patients, 376 were included in the derivation and 162 in the internal validation cohort from the institute; 152 patients from other centers were included in the external validation cohort. The model using the following variables: T2DM duration<5 years (p=0.003), absence of coronary artery disease (p=0.05), absence of diabetic retinopathy (p=0.001), presence of oliguria (p=0.02), acute rise in serum creatinine (p< 0.001), and low serum complement-C3 level (p=0.001) predicted NDKD by multivariate regression analysis. A nomogram was developed to predict the probability of NDKD based on these individual variables, and the model performance was assessed. The model performed robustly with an AUC-ROC of 0.869(95%CI:0.805-0.933) on internal validation and 0.883(95%CI:0.830-0.937) on multicentric external validation. Conclusion The clinical and laboratory parameter-based non-invasive prediction model robustly predicted NDKD among T2DM patients with renal dysfunction.