A non-invasive predictive model for identifying non-diabetic kidney disease in type 2 diabetes mellitus: development and multicenter validation.

Abstract

Background: This study aimed to develop a non-invasive, simple, and rapid predictive model for identifying non-diabetic kidney disease (NDKD) in patients with type 2 diabetes mellitus (T2DM).

Methods: We performed a retrospective analysis of clinical data from 117 T2DM patients who underwent renal biopsy at a single medical institution between 2017 and 2022; candidate variables were first prioritized based on clinical relevance, followed by the construction of a predictive framework using logistic regression. Dubbed the RICH model, the final framework integrated four key parameters: red blood cell (RBC) count, immunoglobulin A (IgA) level, cystatin C-derived estimated glomerular filtration rate (eGFR_2), and glycated hemoglobin A1c (HbA1c).

Results: External validation was conducted across three independent centers involving 299 T2DM patients (2018-2024), achieving area under the receiver operating characteristic curve (AUC-ROC) values of 0.755, 0.764, and 0.755, which complemented the internal validation AUC-ROC of 0.847; at an optimal threshold probability of 0.559, approximately 20% of patients obtained clinical net benefit from the model, and notably, applying the RICH model for early NDKD screening has the potential to reduce the renal biopsy rate by 42.05%.

Conclusions: The RICH model exhibits robust performance in predicting NDKD among T2DM patients with renal impairment, providing a practical tool for clinical decision-making.