Development and validation of a multivariable prediction model for non-invasive discrimination between diabetic and non-diabetic kidney disease in type 2 diabetes: a clinical nomogram

TL;DR

This study created a non-invasive tool to help doctors distinguish between two types of kidney disease in type 2 diabetes patients.

Contribution

A new clinical nomogram was developed to differentiate diabetic and non-diabetic kidney disease using non-invasive variables.

Findings

A five-variable model showed good discrimination and acceptable calibration for DKD and NDKD.

The model included diabetes duration, retinopathy, blood pressure, glucose, and hemoglobin levels.

Decision curve analysis indicated the model's potential for clinical use.

Abstract

This study aimed to develop a non-invasive diagnostic model to differentiate diabetic kidney disease (DKD) from non-diabetic kidney disease (NDKD) in type 2 diabetes mellitus (T2DM) patients with renal insufficiency.

We conducted a retrospective, biopsy-based study of diabetic patients with kidney dysfunction between July 2018 and August 2023. Patients were randomly split into training and validation cohorts (7:3). A multivariable logistic regression model based on routinely available, non-invasive clinical variables was developed and internally validated. Discrimination and calibration were evaluated in both cohorts.

A total of 507 patients were enrolled: 171 with DKD, 260 with NDKD, and 76 with concurrent DKD and NDKD. A five-variable model incorporating diabetes duration, diabetic retinopathy, systolic blood pressure, fasting plasma glucose, and hemoglobin levels demonstrated good discrimination and acceptable calibration in both datasets. Decision curve analysis suggested the model’s potential clinical utility. The model was presented as a nomogram.

This nomogram may support non-invasive differential diagnosis between DKD and NDKD in T2DM patients with kidney injury, thereby informing clinical decision-making.

Figures