Associations of Retinal Vessel Geometry and Optical Coherence Tomography Angiography Metrics With Choroidal Metrics in Diabetic Retinopathy

Abstract

Purpose

To elucidate the mechanism underlying changes in choroidal metrics (choroidal thickness [CT], choroidal vascularity index [CVI], and choriocapillaris [CC] flow deficit [FD]) observed in diabetic retinopathy (DR) and examine the association of choroidal metrics with both retinal vessel geometry and optical coherence tomography angiography (OCTA) metrics.

Methods

Overall, 133 eyes of 133 patients were analyzed retrospectively. Retinal vessel geometry parameters were assessed using semiautomated software. The OCTA metrics and CT were calculated using automated algorithms provided by the manufacturer, whereas the CVI and CC-FD were calculated using ImageJ software from the binarized choroid B-scan image and the CC slab provided by the manufacturer, respectively. To assess the associations among choroidal metrics, retinal vessel geometry, and OCTA metrics, multivariable regression analyses were performed while controlling for clinical features and DR severity.

Results

In the multivariable linear regression analysis, CT (β = −399.84; P = 0.014) and CVI (β = −2.34; P = 0.021) showed significant associations with the arteriole–venule ratio, which is a ratio of central retinal arteriolar equivalent caliber with respect to central retinal venular equivalent caliber. The CC-FD showed a significant association with the fractal dimension of retinal arteriolar network (β = −2.90; P = 0.040). In contrast, the OCTA metrics showed no significant association with the choroidal metrics.

Conclusions

The CT, CVI, and CC-FD in patients with DR were associated with retinal arteriolar geometry parameters rather than OCTA metrics, which indicates an association between choroidal changes and hemodynamic alterations in retinal arterioles and venules.

The choroid is crucial in supplying oxygen and nutrients to the outer retina, including the photoreceptor cells and RPE, and also in removing metabolic waste products. Furthermore, the choroid is involved in light absorption, aids in thermoregulation, and regulates intraocular pressure via vasomotor mechanisms.¹ Although the retinal blood vessels are conspicuous, the primary blood supply to the retina actually comes from the choroid, because more than 90% of the oxygen delivered to the retina is consumed by the photoreceptors.¹ Therefore, impaired choroidal blood flow can lead to the dysfunction and eventual death of photoreceptors, resulting in vision loss.²

Choroidal abnormalities in patients with diabetes mellitus (DM) have only recently begun to receive attention, largely owing to technical constraints. Advancements in optical coherence tomography (OCT), particularly with enhanced depth imaging (EDI) techniques, have allowed for more precise visualization of the structure and layout of the choroid. Moreover, the introduction of OCT angiography (OCTA), especially with swept-source technologies, has enabled more precise evaluation of blood flow in the choriocapillaris (CC).³ Various studies have noted changes in the choroidal thickness (CT), choroidal vascularity index (CVI), and CC in patients with diabetic retinopathy (DR).^4,5 However, the mechanisms behind these choroidal changes have not yet been elucidated; therefore, the clinical application of CT, CVI, and CC flow deficit (FD) remains limited.

Several studies have suggested strongly that the changes in choroidal and retinal vasculatures due to DM are interdependent. First, significant changes in CT have been reported with varying DR severity.⁶ Additionally, considerable overlap exists between retinal and choroidal neovascularization in the involvement of vasoactive molecules.⁷ Despite these findings, the exact association between the choroidal and retinal vasculatures, based on a quantitative analysis of vascular metrics, has not been evaluated.

Given the well-known retinal vascular changes in DR, assessing their association with choroidal changes could reveal the mechanisms behind choroidal vascular alterations in DR. Therefore, we aimed to evaluate the association between choroidal metrics—such as CT, CVI, and CC FD—with retinal vessel geometry and OCTA metrics in this study.

Methods

Ethical Approval

This study was approved by the Institutional Review Board of Jeju National University Hospital (approval number: 2019-07-002) and adhered to the Declaration of Helsinki. The requirement for informed consent was waived owing to the retrospective nature of the study.

Study Population

This retrospective, cross-sectional study was conducted at the Jeju National University Hospital, Jeju-si, Jeju-do, Republic of Korea, and included all consecutive patients referred for DR screening who underwent digital retinal photography and OCTA scans from March 2019 to April 2020. Most of the patients in this study population were included in a previous report.⁸

This study included patients whose conditions ranged from no DR to different stages of DR, including nonproliferative DR (NPDR) and proliferative DR (PDR). The exclusion criteria included a history of previous laser photocoagulation or intravitreal injections or any intraocular surgery, with the exception of cataract surgery. In addition, patients with concurrent retinal diseases other than DR, media opacities that could interfere with fundus photography or OCTA quality, and substandard image quality were excluded.

Ophthalmic Examination

Each patient underwent a comprehensive ophthalmological examination for DR screening, including fundus photography, OCT, and OCTA. Fundus findings were categorized into five stages according to the International Clinical Classification of DR.⁹ In cases where fundus photography, OCT, and OCTA were performed on both eyes of a patient; the eye with higher quality images was included in the analysis.

Retinal Vessel Geometry

Fundus photography was performed using a Visucam NM/FA fundus camera (Carl Zeiss Meditec, Inc., Dublin, CA, USA) in 45° mode centered on the optic disc. Analysis of the retinal vessel geometry was conducted by two independent, trained evaluators (D.J.M. and J.M.C.) using semiautomated software (Singapore “I” Vessel Assessment [SIVA], cloud-based version, National University of Singapore, Singapore), as described in our previous report.⁸ In brief, the software automatically delineated and cataloged the retinal vessels (both arterioles and venules) of the retinal zone, extending from the second to the fifth optic disc away from the optic disc center, producing a skeletonized representation of the retinal vasculature. Manual adjustments were made by the evaluators to correct any errors in the optic disc identification, vascular tracing, or vessel classification. Then, the SIVA software autonomously calculated the retinal vessel geometric parameters, as detailed in Supplementary Table S1.¹⁰

OCTA Metrics and Choroidal Metrics

Structural OCT and OCTA scans were obtained using a PLEX Elite 9000 (Carl Zeiss Meditec, Jena, Germany) with a 6 × 6-mm field of view centered on the fovea. The OCT and OCTA scans of the same eye used for the retinal geometry analysis were included.

The OCTA metrics analyzed in this study were the vessel area density (VAD), representing the percentage of perfused vascular area within a given unit area, scaled from 0.00% (indicating no perfusion) to 100.00% (indicating full perfusion), as assessed in both the superficial (VAD_SCP) and deep capillary plexuses (VAD_DCP); vessel length density (VLD) was used to quantify the length of perfused vessels per unit area in the measured region (mm⁻¹) and was analyzed in both the superficial (VLD_SCP) and deep capillary plexuses (VLD_DCP); the foveal avascular zone (FAZ) circularity index, indicating the compactness compared to a perfect circle based on the perimeter and area, with values spanning from 0.00 (noncircular) to 1.00 (perfectly circular); the FAZ size, calculated in square millimeters; and CT, calculated as the average thickness of the central Early Treatment Diabetic Retinopathy Study subfield, a 1-mm circular area centered at the fovea. The OCTA metrics and CT were automatically generated using the ARI Network Test Algorithms (Macular Density Algorithm v0.7.2 and Choroid Quantification v20220224-B, respectively) supplied by the manufacturer (available at https://arinetworkhub.com/). The eyes with segmentation errors were excluded from the analysis after reviewing the automatically generated segmentation lines display and angiographic en face images.

The CVI is the ratio of the luminal volume of choroidal vessels to the total choroidal volume, expressed as a percentage. We manually measured the CVI of the subfoveal choroidal area, centered at the fovea with a width of 1.5 mm using ImageJ software (National Institutes of Health, Bethesda, MD). Figure A shows a B-scan image of the choroid and retina. Figure B shows a binarized B-scan image of the choroid automatically generated using the ARI Network (white, luminal areas; black, stromal areas).

Figure.

Representative optical coherent tomography angiography images. (A) B-scan image of the choroid and retina. (B) Automatically generated binarized image of the choroid using the ARI Network. (C) Automatically generated projection artifact compensated by the CC slab image ranging from the Bruch membrane (BM) to 20 µm beneath the BM using the ARI Network. (D) Binarized CC slab image using the Niblack's auto-local threshold with the Image J software (National Institutes of Health).

The projection artifact compensated CC slab images, ranging from the Bruch membrane to 20 µm beneath it, were automatically generated by the ARI Network (Fig. C). CC-FD represents the percentage of the nonvascular area within a given unit area, scaled from 0.00% (indicating fully perfused) to 100.00% (indicating no perfusion). The CC slab images were binarized using Niblack's auto-local threshold with ImageJ software to distinguish the vascular area (white area) from the nonvascular area (black area) (Fig. D). The proportion of nonvascular areas within a given unit area was then calculated as a percentage.

Statistical Analysis

Data are presented as mean ± SD or as frequency and percentage. To increase statistical power, the DR severity levels of mild and moderate NPDR were combined into one group, and severe NPDR and PDR were grouped together for analysis and comparison. Comparisons based on the DR severity were conducted using ANOVA and post hoc Tukey tests. Categorical variables were compared using the χ² test.

The analysis used the average value of the retinal vessel geometry parameters assessed by the two evaluators. To understand the mechanisms underlying the changes in choroidal metrics related to DR, the associations with vessel geometry parameters and OCTA metrics were evaluated using linear regression analyses. Preliminary univariate linear regression analyses revealed potential associations among choroidal metrics, retinal vessel geometry, and OCTA metrics. Variables with a P value of less than 0.05 were investigated further using multivariate linear regression models adjusted for clinical characteristics significantly associated with choroidal metrics.

Statistical significance was set at a P value of less than 0.05, and statistical analyses were conducted using IBM SPSS Statistics for Windows (version 22.0; IBM Corp., Armonk, NY).

Results

Overall, 133 eyes of 133 patients were included in the study. Based on the International Clinical Disease Severity Scale for DR, eyes were categorized as having no DR (n = 48 [36.09%]), mild/moderate NPDR (n = 49 [36.84%]), or severe NPDR/PDR (n = 36 [27.07%]).

The clinical features of patients with different levels of DR severity are shown in Supplementary Table S2. Patients with mild/moderate NPDR exhibited lower hemoglobin A1c (HbA1c) levels than those without DR (P = 0.019). Patients with severe NPDR/PDR had a longer DM duration (P = 0.027) and greater central macular thickness (P = 0.007) than those without DR.

Intergrader reliability estimates for retinal vessel geometry parameters are presented in Supplementary Table S3. The intergrader reliability was strong across all retinal vessel geometry parameters, with intraclass correlation coefficients of 0.80 or greater. The mean retinal vessel geometry parameters at each DR severity level evaluated by the two graders are listed in Supplementary Table S4. Patients with severe NPDR/PDR showed a smaller arteriole–venule ratio (AVR) than those with mild/moderate NPDR (P = 0.046). In addition, patients with severe NPDR/PDR demonstrated a smaller fractal dimension of arteriolar network (FDa) values than those without DR (P = 0.012) and those with mild/moderate NPDR (P = 0.011).

The OCTA metrics for each DR severity level are presented in Supplementary Table S5. Patients with severe NPDR/PDR exhibited a lower VAD_DCP (P = 0.008) and VLD_DCP (P = 0.005) than those with no DR. In addition, compared with patients without DR and patients with mild/moderate NPDR, patients with severe NPDR/PDR demonstrated a smaller VAD_SCP (P < 0.001 and P = 0.003, respectively), VLD_SCP (P = 0.001 and 0.007, respectively), and FAZ circularity (P = 0.043 and 0.027, respectively).

The choroidal metrics of patients at each DR severity level are shown in Table 1. Patients with severe NPDR/PDR exhibited a larger CT than those without DR (P = 0.038). In addition, compared with patients without DR and patients with mild/moderate NPDR, patients with severe NPDR/PDR demonstrated significant increases in CVI (P = 0.034 and 0.040, respectively) and CC-FD (P < 0.001 and P = 0.038, respectively).

Table 1.

Data of Comparison of Choroidal Metrics According to DR Severity

Choroidal Metrics	No DR	Mild/Moderate NPDR	Severe NPDR/PDR	P Value*
CT (µm)	270.71 ± 88.77	303.86 ± 100.68	327.87 ± 126.92	0.045
CVI (%)	59.21 ± 3.74	59.27 ± 2.91	61.18 ± 3.94	0.022
CC FDs (%)	56.31 ± 0.88	56.65 ± 0.99	57.18 ± 1.03	<0.001

^*One-way ANOVA test.

Associations Between Choroidal Metrics, Retinal Vessel Geometry Parameters, and OCTA Metrics and Clinical Characteristics

The associations between CT, CVI, and CC-FD and clinical characteristics are shown in Table 2. Age was associated with CT (β = −2.30; P < 0.001), CVI (β = −0.06; P = 0.005), and CC-FD (β = 0.04; P < 0.001). The presence of hypertension associated with CC-FD (β = 0.36; P = 0.041). DM duration was associated with CT (β = −3.10; P = 0.004) and CC-FD (β = 0.05; P < 0.001), whereas the HbA1c level was associated with CVI (β = −0.42; P < 0.001). DR severity was associated with CT (β = 22.82; P = 0.013), CVI (β = 0.93; P = 0.018), and CC-FD (β = 0.43; P < 0.001).

Table 2.

Data Depicting the Relationship Between Clinical Characteristics and Choroidal Metrics

	Choroidal Metrics
	CT		CVI		CC FD
Clinical Characteristics	β	P Value *	β	P Value*	β	P Value *
Age (years)	−2.30	<0.001	−0.06	0.005	0.04	<0.001
Male	−36.41	0.063	−0.31	0.641	0.02	0.900
Hypertension	−23.95	0.200	−0.67	0.291	0.36	0.041
Duration of DM (years)	−3.10	0.004	−0.03	0.363	0.05	<0.001
HbA1c (%)	−2.08	0.533	−0.42	<0.001	0.02	0.654
DR severity	28.82	0.013	0.93	0.018	0.43	<0.001
Central macular thickness (µm)	0.09	0.762	0.01	0.237	1.17 × 10−4	0.965

^*One-way ANOVA test.

In the univariable linear regression analysis of CT with the retinal vessel geometric parameters, the AVR (β = −442.87; P = 0.008), fractal dimension of the venular network (β = 396.79; P = 0.043), and arteriolar junctional exponent deviation (β = 52.65; P = 0.031) showed statistically significant associations (Table 3). The multivariate linear regression analysis, adjusted for age, DM duration, and DR severity showed statistically significant results (adjusted R² = 0.215; P < 0.001); the CT was significantly associated with AVR (β = −399.84; P = 0.014). However, in the univariate linear regression analysis, CT was not significantly associated with any of the OCTA metrics (Supplementary Table S6).

Table 3.

Data Depicting the Relationship Between the Retinal Vessel Geometry Parameters and CT

	CT
	Univariable		Multivariable*
Retinal Vessel Geometry	β	P Value	β	P Value
CRAE	0.09	0.280
CRVE	0.08	0.174
AVR	−442.87	0.008	−399.84	0.014
FDa	146.57	0.384
FDv	396.79	0.043	185.79	0.353
TORTa	−1.30 × 105	0.682
TORTv	−1.24 × 105	0.662
BCa	−28.35	0.240
BCv	−12.73	0.709
JEa	52.65	0.031	25.56	0.353
JEv	24.29	0.351

BCa, arteriolar branching coefficient; BCv, venular branching coefficient; CRAE, central retinal arteriolar equivalent caliber; CRVE, central retinal venular equivalent caliber; FDv, fractal dimension of venular network; Jea, junctional exponent deviation for arterioles; JEv, junctional exponent deviation for venules; TORTa, arteriolar curvature tortuosity; TORTv, venular curvature tortuosity.

^*Adjusted for age, duration of diabetes, and DR severity.

In the univariable linear regression analysis of CVI with the retinal vessel geometric parameters, AVR was significantly associated with CVI (β = −13.46; P = 0.018) (Table 4). The multivariate linear regression analysis, adjusted for age, HbA1c level, and DR severity, showed statistically significant results (adjusted R² = 0.206; P < 0.001); the CVI was significantly associated with AVR (β = −2.34; P = 0.021). However, in the univariate linear regression analysis, CT was not significantly associated with any of the OCTA metrics (Supplementary Table S7).

Table 4.

Data Depicting the Relationship Between the Retinal Vessel Geometry Parameters and the CVI

	CVI
	Univariable		Multivariable*
Retinal Vessel Geometry	β	P Value	β	P Value
CRAE	−1.31 × 10−3	0.623
CRVE	−5.52 × 10−4	0.772
AVR	−13.46	0.018	−2.34	0.021
FDa	4.91	0.393
FDv	9.10	0.173
TORTa	4200.11	0.710
TORTv	7396.33	0.442
BCa	0.584	0.501
BCv	−0.03	0.981
JEa	−0.07	0.931
JEv	−0.26	0.764

BCa, arteriolar branching coefficient; BCv, venular branching coefficient; CRAE, central retinal arteriolar equivalent caliber; CRVE, central retinal venular equivalent caliber; FDv, fractal dimension of venular network; JEa, junctional exponent deviation for arterioles; JEv, junctional exponent deviation for venules; TORTa, arteriolar curvature tortuosity; TORTv, venular curvature tortuosity.

^*Adjusted for age, HbA1c level, and DR severity.

In the univariable linear regression analysis of CC-FD with the retinal vessel geometric parameters, FDa (β = −6.85; P < 0.001) was significantly associated with CC-FD (Table 5). The multivariate linear regression analysis, adjusted for age, presence of hypertension, duration of DM, and DR severity, showed statistically significant results (adjusted R² = 0.420; P < 0.001); the FDa was significantly associated with CC-FD (β = −2.90; P = 0.040). In the univariable linear regression analysis of CC-FD with the OCTA metrics, the VAD_SCP (β = −0.15; P < 0.001), VAD_DCP (β = −0.06; P < 0.001), VLD_SCP (β = −0.30; P < 0.001), VLD_DCP (β = −0.12; P < 0.001), FAZ circularity (β = −3.43; P < 0.001), and FAZ size (β = 0.62; P = 0.006) were significantly associated with CC-FD (Table 6). However, in the multivariable linear regression analysis, adjusted for age, presence of hypertension, duration of DM, and DR severity, the OCTA metrics were not significantly associated with CC-FD.

Table 5.

Data Depicting the Relationship Between the Retinal Vessel Geometry Parameters and CC VAD

	CC FDs
	Univariable		Multivariable*
Retinal Vessel Geometry	β	P Value	β	P Value
CRAE	−5.12 × 10−4	0.491
CRVE	−4.15 × 10−4	0.432
AVR	−1.10	0.495
FDa	−6.85	<0.001	−2.90	0.040
FDv	−3.13	0.095
TORTa	163.91	0.958
TORTv	2686.50	0.315
BCa	0.20	0.397
BCv	0.42	0.193
JEa	−0.33	0.152
JEv	−0.05	0.855

BCa, arteriolar branching coefficient; BCv, venular branching coefficient; CRAE, central retinal arteriolar equivalent caliber; CRVE, central retinal venular equivalent caliber; FDv, fractal dimension of venular network; JEa, junctional exponent deviation for arterioles; JEv, junctional exponent deviation for venules; TORTa, arteriolar curvature tortuosity; TORTv, venular curvature tortuosity.

^*Adjusted for age, presence of hypertension, duration of diabetes, and DR severity.

Table 6.

Data Depicting the Relationship Between the OCTA Metrics and CC VAD

		CC FDs
		Univariable		Multivariable*
OCTA Metrics		β	P Value	β	P Value
VAD (%)	SCP	−0.15	<0.001	−0.12	0.255
	DCP	−0.06	<0.001	0.06	0.744
VLD (mm−1)	SCP	−0.30	<0.001	−0.08	0.757
	DCP	−0.12	<0.001	−0.10	0.800
FAZ	Circularity	−3.43	<0.001	−1.31	0.125
	Size (mm2)	0.62	0.006	0.12	0.574

DCP, deep capillary plexus; SCP, superficial capillary plexus.

^*Adjusted for age, presence of hypertension, duration of diabetes, and DR severity.

Discussion

Changes in the choroidal metrics in DR were significantly associated with retinal arteriolar geometry parameters but not with OCTA metrics. CT and CVI were associated with the AVR, whereas CC-FD was associated with FDa. To the best of our knowledge, this study is the first to evaluate the association between choroidal metrics and both retinal vessel geometry and OCTA metrics to elucidate the mechanism underlying choroidal changes observed in DR.

Choroidal changes in patients with DR is referred to as diabetic choroidopathy (DC). The major findings of DC are the microaneurysms, dilation or narrowing of vascular lumen, increased vascular tortuosity, vascular drop-out and areas of vascular nonperfusion, and choroidal neovascularization. These pathological changes primarily affect the CC, although larger choroidal vessels may also be affected.¹¹

Before the introduction of EDI-OCT or OCTA, the evaluation of DC was limited because it relied on either histological studies of enucleated eyes or indocyanine green angiography. EDI-OCT and OCTA offer several significant advantages for evaluating DC, including noninvasiveness, the ability to conduct repeated in vivo imaging without the need for dye, and cost effectiveness. Moreover, these techniques are quick to perform, making them ideal for routine clinical use. The choroidal metrics that can be evaluated using OCT and OCTA are CT, CVI, and CC-FD.

CT is a sensitive biomarker for various ocular and systemic diseases.^12,13 Eyes with pachychoroid spectrum diseases or Vogt–Koyanagi–Harada disease show increased CT. In contrast, eyes with AMD, pathological myopia, retinal dystrophies, or idiopathic macular holes show decreased CT.¹² However, CT solely reflects the entirety of the choroidal structure without distinguishing between its stromal and vascular components, and does not reflect CC status. The CVI measures the choroidal vasculature status by calculating the luminal and stromal choroidal areas.¹⁴ The CVI potentially may act as a marker of various ocular diseases, wherein eyes with acute central serous chorioretinopathy and Vogt–Koyanagi–Harada disease show an increased CVI, and eyes with AMD show a decreased CVI.¹⁵ Nevertheless, CVI primarily reflects changes in large choroidal vessels and is limited by its inability to assess changes in the CC. The en face images of the CC can be subjected to image processing to quantitatively evaluate parameters, such as vessel density and CC-FD.¹⁶

We found that eyes with severe NPDR/PDR exhibited significantly increased CT compared with those without DR, which is consistent with previous studies.¹⁷ The increased CT observed in severe DR is understood to be primarily attributable to elevated intraocular levels of VEGF, which induce various changes in the choroid, such as dilation of the choroidal vessels, increased blood flow, and enhanced vascular permeability.¹⁷ This is further supported by our findings, which demonstrated a significantly higher CVI in eyes with severe NPDR/PDR compared with those with no DR or mild/moderate NPDR. The increased CVI is primarily due to the dilation of the large and medium choroidal vessels in the Haller's and Sattler's layers, which is likely a result of elevated intraocular VEGF levels. Reports of reduced CT and CVI after intravitreal anti-VEGF injections suggest that CT and CVI depend on the VEGF levels.^18–20

Some studies have reported that CT and CVI either remain unchanged or even decrease with increasing DR severity.^4,17 The discrepancies in these findings could be influenced by factors such as the patient's DM management, whether they received intravitreal anti-VEGF injection or panretinal photocoagulation treatment, or the presence of diabetic macular edema (DME). In our study, we included newly referred patients screened for DR with a short history of DM treatment and excluded those who previously received anti-VEGF or panretinal photocoagulation treatment. This enhanced the reliability of our findings regarding the association between DR severity and changes in CT and CVI.

We demonstrated that a decrease in AVR was associated with an increase in CT and CVI. AVR, a composite measure reflecting retinal arteriolar narrowing or retinal venular widening, has been studied extensively for its association with various systemic conditions, such as hypertension, DM, carotid artery stiffness, coronary heart disease, and cerebral infarction.²¹ However, the association between AVR and intraocular VEGF levels has not been studied. In patients with DM, narrower retinal arterioles have been reported to be associated with macular ischemia²² and systemic ischemic complications, such as lower limb amputation,²³ and peripheral neuropathy.²⁴ Intravitreal injections of VEGF resulted in dilatation of retinal venules in cynomolgus monkeys.²⁵ Taking these findings into account, there is a significant possibility that the decrease in AVR due to retinal arteriolar narrowing and venular dilation may be associated with retinal ischemia, followed by an increase in intraocular VEGF levels. Increased intraocular VEGF levels are likely to induce the dilation of choroidal vessels, increase choroidal blood flow and vascular permeability, leading to an increase in CVI and CT. Meanwhile, OCTA metrics reportedly do not show significant changes before and after intravitreal VEGF injections.^26,27 Therefore, it is likely that no association between OCTA metrics and VEGF levels existed, which explains why no association between OCTA metrics and CT and CVI was observed.

An increase in CC-FD is a well-known feature of DC and is correlated with DR severity.²⁸ The mechanisms underlying the increased CC-FD in DC remain unclear but may involve impairment of the CC or the occurrence of signal voids due to decreased blood flow. Elevated CC-FD was not associated with reduced VAD, VLD, or an enlarged FAZ, but it was associated with a decrease in FDa. This finding suggests that, although the increase in CC-FD is driven by a mechanism distinct from retinal capillary closure, it may share a common pathway with the decrease in FDa. A decrease in central retinal arteriolar equivalent in severe DR has been reported in several studies,^8,29 which can lead to retinal ischemia, followed by an increase in intraocular VEGF levels, as described elsewhere in this article. Consequently, the dilation of large choroidal vessels by VEGF may induce a thinning of the CC, resulting in increased CC-FD, as observed in pachychoroid spectrum diseases.³⁰ This potential mechanism may explain the association between increased CC-FD and decreased FDa, but further studies are needed to validate this hypothesis.

The clinical significance of DC has been under-recognized historically owing to limitations in evaluation techniques. However, recent advances in imaging technology, enabling more precise and accessible assessments, are likely to stimulate further research into DC aimed at deepening our understanding of its clinical implications. The loss of CC in DC can lead to hypoxia in the affected areas of the choroid and the overlying RPE. In response, the RPE upregulates VEGF production, which in turn stimulates angiogenesis.³¹ Additionally, CC perfusion has been identified as the vascular variable that most significantly affects photoreceptor structure, as the outer retina primarily relies on diffusion from the choroidal circulation to meet its oxygen and nutrient demands.³² Moreover, CC-FD has been reported as independently associated with DR progression and the development of DME in patients with type 2 DM.³³ Therefore, even without OCTA, a decreased FDa observed on fundus photography, which suggests increased CC-FD, may indicate an increased risk of DR progression, neovascularization, DME occurrence, and photoreceptor damage, potentially leading to poor visual outcomes.

Several studies have shown that increased CVI and CT can predict a favorable response to anti-VEGF therapy in DME.^34,35 Additionally, an increased CT has been identified as a new indicator for monitoring DME recurrence.³⁶ This is due to VEGF-mediated vascular hyperpermeability, which plays a key role in the pathophysiology of DME and contributes to the increase in CVI and CT. Thus, even in the absence of EDI-OCT and OCTA, a decreased AVR observed on fundus photography, which suggests increased CVI and CT, may indicate a favorable response to anti-VEGF treatment in DME and risk of DME recurrence.

Our study focused on the central retina; however, with the advent of ultra-widefield imaging systems, the evaluation of retinal vessel geometry, OCTA metrics, and choroidal metrics in the peripheral retina has become feasible. Histopathological changes in DC are most pronounced in the area between the equator and ora serrata.^37,38 In contrast, CVI changes according to DR severity are most evident in the central macula.³⁹ Therefore, future research should investigate whether differences in field of view influence the associations between retinal vessel geometry, OCTA metrics, and choroidal metrics when using ultra-widefield imaging systems to encompass more of the peripheral retina.

Our study had some limitations. First, refractive error, axial length, smoking status, and diurnal variation, which may influence retinal vessel geometry, OCTA metrics, and choroidal metrics, could not be evaluated. Second, the participants without DM were excluded from the analyses owing to the paucity of OCTA images. Even in patients having DM without DR, retinal vessel geometry, OCTA metrics, and choroidal metrics exhibit less pronounced changes but follow a similar pattern to those with noticeable DR.⁴⁰ To understand the early changes and pathogenesis of DR and DC, individuals without DM are more suitable as controls. To validate our findings, future prospective and longitudinal studies should include healthy controls and consider other possible confounding factors. Third, although the association between choroidal metrics and retinal vessel geometry identified in this study was statistically significant, it demonstrated low-to-moderate explanatory power. This finding indicates that further research is needed to explore additional variables that could better explain the relationship between choroidal metrics and retinal vessel geometry, thereby enhancing the explanatory power of the findings. Fourth, the automated algorithm for OCTA metrics analysis has not yet been validated fully. However, a recent study demonstrated that automated FAZ area determination using the ARI network is feasible and yields results comparable with manual measurements, with perfect intrascan reproducibility and excellent interscan reproducibility, thereby affirming the reliability of the ARI Network's automated analysis algorithm.⁴¹ Furthermore, the macular density algorithm of the ARI network has been cited in approximately 100 peer-reviewed publications, providing substantial evidence of its validation and reliability in the field. Similarly, retinal vessel geometry analysis using SIVA software has been used in more than 400 peer-reviewed publications, further confirming its validity and reliability in clinical research. Last, common comorbidities associated with DR, such as DME, macular ischemia, epiretinal membrane, and vitreomacular traction, can cause vessel misidentification, segmentation errors, and noise artifacts in OCTA imaging.⁴² Segmentation errors are the most common,⁴³ and most studies have attempted manual correction. In contrast, CC segmentation has fewer errors because it is based on Bruch's membrane, which remains relatively intact even in DME. Manual correction is not feasible with the PLEX Elite 9000 and ARI Network test algorithms. Therefore, eyes with confirmed segmentation errors were excluded from this study. Currently, there are no standardized clinical protocols for managing OCTA artifacts, highlighting the need for future updates to OCTA devices and software to address these issues.

In this study, we demonstrated that a decreased FDa is associated with increased CT and CVI, likely mediated by elevated intraocular VEGF levels, which may indicate a higher risk of DR progression and poor visual outcomes. Additionally, a decreased AVR is associated with increased CC-FD, also mediated by elevated intraocular VEGF levels, which may suggest a favorable response to anti-VEGF treatment and potential recurrence in DME. Further research is needed to strengthen the evidence for our findings, which will enhance our understanding of DR and DC.