In this single-center, cross-sectional study, we recruited adult patients diagnosed with T2DM without diabetic retinopathy. This study was conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board of the Medical Faculty of Mannheim (ID number: 2023 − 666). Written informed consent was obtained before participation.
We screened 123 type 2 diabetic patients who presented to the ophthalmology or endocrinology clinic of a tertiary referral center. The study included patients with T2DM and without diabetic retinopathy in any eye, after review of 5-field color fundus photography. Optical coherence tomography (OCT), and OCT angiography (OCTA) were also performed. Exclusion criteria were: low-quality OCT and OCTA images (defined as a signal-to-noise ratio (Q-value) < 30dB or with significant image artifacts: attenuation, striping, defocus or projection), presence of any retinopathy, glaucoma or significant media opacities. When both eyes of the same patient met inclusion criteria, the eye with the best image quality defined by a retina specialist was selected.
We excluded 31 patients due to presence of clinically visible diabetic retinopathy seen on fundus examination, 5 patients for having concomitant age-related macular degeneration and 5 additional patients due to poor image quality in both eyes, leaving 82 patients for further analysis.
Primary outcome measures were duration of the diabetes (in years) and its association with vascular OCTA parameters, including vessel density (VD), vessel length density (VLD), choriocapillaris flow deficit area (CC FD%) and foveal avascular zone (FAZ) circularity. Exploratory outcomes included associations with arterial hypertension, age and sex. Patients taking antihypertensive medication were defined as having arterial hypertension.
Image acquisition
We acquired macular OCT images with a Spectral-Domain OCT Spectralis HRA-OCT3 (Heidelberg Engineering, Heidelberg, Germany), with 18 averaged images by automatic real time function, a 122 spacing between B-scans and at 85 kHz scan speed.
OCTA imaging protocol included a 20 × 20º macular scan at 85 kHz A-scan rate and a spacing of 11 microns between B-scans.
Image processing and analysis
En face projections of OCTA images at the superficial (SVP), intermediate (ICP) and deep (DVP) vascular plexus and the choriocapillaris (CC) were exported in a 1:1 pixel scale and imported into ImageJ® (version 2.14.0/1.54f, National Institutes of Health, Bethesda, MD, USA). The segmentation boundaries used to define the different retinal plexuses were those preset by the HEYEX software (Heidelberg Engineering, Heidelberg, Germany). The SVP contained the ganglion cell layer and the inner plexiform layer; the ICP was situated between inner plexiform layer and inner nuclear layer, the DVP between inner and outer plexiform layer and the CC 10 μm below the Bruch’s membrane.
Using imageJ®, images were first binarized with the Phansalkar Auto Local Threshold method (radius = 15) and an ETDRS grid was added centered on fovea [6, 7]. Vessel density (VD, defined as white pixels/image area, %) was calculated in the binarized images in both the SVP and DVP in the nasal, superior, temporal and inferior sectors of the ETDRS grid (Fig. 1). Subsequently, images were skeletonized to calculate vessel length density in the same sectors (VLD, defined as vessel length per unit area, mm− 1).
En face OCTA scan of the superficial vascular plexus after binarization and skeletonization. The nasal, superior, temporal and inferior sectors of the ETDRS grid centered on the foveal avascular zone we included in the analysis are depicted
CC flow deficit percentage (CC FD, %) was automatically calculated from normalized and then binarized images of the choriocapillaris slab, using the Phansalkar’s local thresholding method, as previously described [8]. CC FD% was then calculated as area without flow signal/total area, using the Analyze Particles tool. We excluded those areas < 24microns as this is equivalent to the normal intercapillary distance [9]. CC FD% was calculated in the same 4 sectors of the ETDRS grid (nasal, superior, temporal and inferior), centered on the same coordinates as the corresponding SVP and DVP images.
The area (mm2) and perimeter (mm) of the FAZ were measured manually using the freehand selection tool in en face images from the ICP, as previous authors have reported a slightly better repeatability of FAZ measurements on this plexus [10]. The circularity of the FAZ was then calculated using the following formula: 4π*area/perimeter2 (unitless), as previously described. FAZ circularity values range from 0 to 1, with a value of 1 representing a perfect circle. We analyzed only FAZ circularity because it showed less interindividual variability than FAZ area and perimeter [11, 12].
Peripheral areas were not included in the analysis to avoid bias due to image attenuation.
Statistical analysis
Reported descriptive statistics include mean values and standard deviation. The ANOVA test was used to compare OCTA metrics between the different areas of the retina studied after assessing the normality of the distribution using the Shapiro-Wilk test.
Linear regression was used to study association between duration of the disease (dependent variable) and OCTA vascular metrics: for each of the retinal sectors (nasal, superior, temporal and inferior) including either VD, VLD, FAZc and CC FD%.
For each variable showing significant association with disease duration in the univariate analysis, a multivariate analysis was performed to adjust the model for the following confounders: age, sex, Hb1Ac and the presence of arterial hypertension.
All analysis were performed using GraphPad Prism® (version 10.2.1).
Leave a Reply