Prediction of the Short-Term Risk of New-Onset Renal Dysfunction in Patients with Type 2 Diabetes: A Longitudinal Observational Study

Abstract

Background. Studies in the past decade have reported many novel biomarkers for predicting the new-onset or progression risk of renal dysfunction in patients with type 2 diabetes (T2D) based on the genomic, metabolomic, and proteomic technologies. These novel predictive markers, however, are difficult to be widely used in clinical practice over the short term due to their high technology content, instability, and high cost. This study was aimed at evaluating the associations of clinical features and six traditional renal markers with the short-term risk of new-onset renal dysfunction in patients with T2D. Methods. This study involved 213 participants with T2D and normal renal function at baseline. The baseline levels of the albumin-to-creatinine ratio (ACR), estimated glomerular filtration rate (eGFR), alpha-1-microglobulin-to-creatinine ratio (A1MCR), neutrophil gelatinase-associated lipocalin-to-creatinine ratio, transferrin-to-creatinine ratio (UTRF/Cr), and retinol-binding protein-to-creatinine ratio (URBP/Cr) were analyzed. Multivariate logistic models were established and validated. Results. During the two-year follow-up period, 23.01% participants progressed to renal dysfunction. The basal levels of ACR, A1MCR, UTRF/Cr, and URBP/Cr were the independent risk factors of new-onset renal dysfunction ( $P<0.05$ ). Several logistic models incorporating clinical characteristics and these renal markers were constructed for predicting the short-term risk of new-onset renal dysfunction. Comparatively, the model including age, glycated hemoglobin (HbA1c), hypertension, ACR, A1MCR, UTRF/Cr, and URBP/Cr levels at baseline had the highest potential ( $\text{C}-\text{index}=0.785$ , $P<0.001$ ). This model was validated using the $K$ -fold cross-validation method; the accuracy was $0.815\pm 0.013$ in training sets and $0.784\pm 0.019$ in validation sets, indicating a good consistency for predicting the new-onset renal dysfunction risk. Finally, a nomogram based on this model was constructed to provide a quantitative tool to assess the individualized risk of short-term new-onset renal dysfunction. Conclusion. The model incorporating these markers and clinical features may have a high potential to predict the short-term risk of new-onset renal dysfunction.