# Changes of retinal microvascular parameters in patients with type 2 diabetes mellitus with or without diabetic kidney disease

## TL;DR

This study shows that retinal microvascular changes in type 2 diabetes patients can help detect early kidney disease risk.

## Contribution

The study introduces retinal microvascular parameters as a non-invasive tool for early detection of diabetic kidney disease.

## Key findings

- Retinal parameters like vessel density and tortuosity differ significantly between T2DM patients with and without DKD.
- VD within the 3-mm foveal avascular zone is an independent risk factor for DKD.
- AI-derived retinal metrics could serve as a non-invasive screening method for kidney disease in T2DM patients.

## Abstract

To investigate the quantitative retinal microvascular parameters in patients with type 2 diabetes mellitus (T2DM) with or without diabetic kidney disease (DKD).

This retrospective study included 117 T2DM patients with DKD (DKD group) and 179 T2DM patients without DKD (non-DKD group). Quantitative retinal microvascular parameters were extracted from fundus photographs by an artificial intelligence (AI) system, including the number of retinal lesions (hard exudates, cotton wool spots, hemorrhagic spots, and microaneurysms), mean branch angle, fractal dimension (FD), mean vessel diameter (MVD), mean arterial diameter (MAD), mean venular diameter (MVeD), arteriole-to-venule ratio (AVR), mean vessel tortuosity (MVT), mean arterial tortuosity (MAT), mean venular tortuosity (MVeT), vessel density (VD), VD within the 3-mm and 5-mm foveal avascular zones (FAZ), and cup-to-disc area ratio (CDR). Systemic parameters were also collected, including systolic blood pressure (SBP), diastolic blood pressure (DBP), fasting plasma glucose (FPG), duration of T2DM, glycated hemoglobin (HbA1c), triglyceride (TG), total cholesterol (CHOL), bicarbonate (HCO3), blood urea nitrogen (BUN), uric acid (UA), creatinine, cystatin C, and estimated glomerular filtration rate (eGFR). Group comparisons were performed using t-test or Mann-Whitney test. Univariable and multivariable logistic regression analyses were used to determine the parameters associated with DKD.

Multiple retinal parameters were significantly different between the two groups, including numbers of hard exudates, cotton wool spots, microaneurysms, hemorrhagic spots, FD, MVD, MVeD, VD, and VD within the 3-mm and 5-mm FAZ (all P < 0.05). Univariable logistic regression analysis demonstrated significant associations between DKD risk and the following variables: SBP, BUN, HCO3, eGFR, creatinine, UA, cystatin C, duration of T2DM, staging of diabetic retinopathy (DR), number of cotton wool spots, microaneurysms, FD, MVD, MVeD, VD, and VD within 3-mm and 5-mm FAZ (all P < 0.05). Multivariable logistic regression analysis found that age (OR = 0.971, 95% CI: 0.948–0.994, P = 0.013), SBP (OR = 1.015, 95% CI: 1.002–1.029, P = 0.027), duration of T2DM (OR = 1.056, 95% CI: 1.019-1.095, P = 0.003), staging of DR (OR = 1.287, 95% CI: 1.064-1.557, P = 0.009), and VD within 3-mm radius of FAZ (OR = 0.000, 95% CI: 0.000-0.000, P < 0.001), were independent risk factors of DKD.

Quantitative retinal microvascular parameters derived from fundus photographs show potential for identifying T2DM patients at risk of DKD, supporting the future use of retinal parameters as a non-invasive tool for early detection of renal impairment in patients with T2DM.