Early and Late Treatment Influence on Chorioretinal Microvasculature in Vogt–Koyanagi–Harada Patients Using Optical Coherence Tomography Angiography

Abstract

Purpose: To study the impact of early and late treatment on chorioretinal microvasculature in Vogt–Koyanagi–Harada (VKH) disease using optical coherence tomography angiography (OCTA).

Methods: A total of 103 patients with VKH disease were divided into early (group 1, starting treatment within 2 months after disease onset) and late (group 2, starting treatment 2 months after disease onset) treatment groups. Flow area (FA) and vessel density (VD) of the retinal superficial vascular complex (SVC) and deep vascular complex (DVC), FA of the choriocapillaris, three-dimensional choroidal vascular volume (CVV), and choroidal vascularity index (CVI) were analyzed and compared to 103 healthy individuals. The relationship between the final best-corrected visual acuity (BCVA) and the aforementioned parameters was also analyzed.

Results: FA of the SVC (all P < 0.05, except 0–1 mm P = 0.087), DVC (all P < 0.05), choriocapillaris (1–2.5 mm P = 0.033), and CVV (all P < 0.05) were lower in group 2 as compared to group 1. Compared to healthy controls, FA of the SVC (all P < 0.001, except 0–1 mm P = 0.104) and DVC (all P < 0.05), VD of the SVC (1–2.5 mm P = 0.001) and DVC (1–5 mm P = 0.003, 2.5–5 mm P < 0.001), FA of the choriocapillaris (all P < 0.05), and CVV (total area P = 0.049, 1–5 mm P = 0.045, 2.5–5 mm P = 0.041) were lower in group 2, while FA (all P < 0.05, except 0–1 mm P = 0.925) and VD (1–5 mm P = 0.003, 2.5–5 mm P = 0.004) of the DVC and FA of the choriocapillaris (total area P = 0.007, 0–1 mm P < 0.001, 1–2.5 mm P = 0.007) were lower in group 1. There was no significant difference concerning CVI among groups (all P > 0.05). FA of the SVC, DVC, and choriocapillaris and VD of DVC and CVI were negatively associated with the final logarithm of the minimum angle of resolution BCVA.

Conclusions: Patients with VKH disease who are treated within 2 months of disease onset showed a better chorioretinal microvascular outcome as defined by OCTA compared to those treated late.

Translational Relevance: Our study employs OCTA to design three-dimensional metrics for the retina and choroid, bridging the gap between traditional two-dimensional OCTA findings and enhanced clinical outcomes for patients with VKH disease.

Introduction

Vogt–Koyanagi–Harada (VKH) disease is a major noninfectious uveitis entity that is more prevalent in middle-aged Asians than in Caucasians.¹ As a multisystem autoimmune disorder caused by a T-cell–dependent response against melanocytes, VKH disease commonly manifests with bilateral panuveitis leading to visual impairment, as well as skin and auditory signs.² Pathologic studies revealed that the choroid, which contains a dense population of melanocytes, is the primary target of inflammation in VKH disease, whereas the retina is involved later, as a possible bystander.³ Chorioretinal involvement during VKH disease can be visualized using indocyanine green angiography and fundus fluorescein angiography. These methods are both invasive modalities with a potential risk of contrast medium allergy and cannot quantify the alterations of the retinal and choroidal microvasculature. The introduction of swept-source–optical coherence tomography angiography (SS-OCTA) has revolutionized the analysis of the fundus. Compared with spectral-domain optical coherence tomography (SD-OCT), SS-OCTA offers several advantages, including deeper penetration into the choroid, higher speed, and better image resolution, allowing for the detailed visualization of both the retinal and choroidal vasculature. This noninvasive technique eliminates the risks associated with contrast agents and provides a comprehensive evaluation of microvascular changes over time. Studies have demonstrated that medium- and large-sized choroidal vessels play a critical role in the pathophysiology of VKH disease. In particular, larger vessels occupy the majority of the choroidal area. Inflammation of these larger vessels can result in significant structural and functional changes within the choroid, contributing to disease progression.⁴ This inflammation leads to choroidal thickening and subsequent visual impairment due to the disruption of normal choroidal blood flow and subsequent retinal involvement. However, studies on VKH disease utilizing SS-OCTA had primarily concentrated on the microcirculation of the retina and the choriocapillaris layer, neglecting the broader distribution and more functionally significant medium- and large-sized vessels of the choroid.^5,6 Due to the more pronounced structure of larger vessels, we can monitor disease activity more accurately, potentially aiding in the early diagnosis and tailored treatment planning for patients with VKH disease. Thus, we used novel artificial intelligence (AI)–based three-dimensional auxiliary SS-OCTA metrics and defined new choroidal-related indicators as applied in other chorioretinopathy.⁷ This approach allows for a more detailed and quantitative analysis of vascular structures, which is crucial for understanding the pathophysiology of VKH disease. Additionally, related research had indicated a reduced retinal vessel density (VD) and normal choriocapillaris flow density in the convalescent VKH disease stage,⁸ whereas other studies found a normal retinal VD and varying degrees of choriocapillaris loss in VKH disease.^6,9,10 The patient cohorts in the studies mentioned above had a variable duration of disease before treatment initiation. Consequently, it is worth exploring whether this variability is the cause of the inconsistent results in the retinal and choroidal microvasculature in VKH disease, which was the subject of the study presented here. 

Methods

Participants

This cross-sectional study was approved by the local institutional review board of the first affiliated hospital of Chongqing Medical University (2023-437). Informed consents were signed by all individuals for collecting the clinical data, and the study followed the statements of the Declaration of Helsinki. We routinely performed OCTA on all patients with VKH disease visiting our uveitis department from September 2019 until May 2023 and selected 103 patients with inactive VKH disease from this cohort. For analysis, we selected the OCTA examination results from a specific time point posttreatment, ensuring that the duration of VKH disease between the two groups remained comparable. Patients with inactive VKH disease all had no detectable anterior chamber cells or keratic precipitates and no inflammatory signs in the fundus according to both the Revised Diagnostic Criteria proposed by the International Nomenclature Committee and our criteria reported previously.^11,12 Patient age, gender, duration of VKH disease, and the time between disease onset and the start of treatment were recorded. We also performed basic ocular examinations, including initial and last follow-up best-corrected visual acuity (BCVA), slit-lamp examination, and intraocular pressure (IOP). The visual acuity was examined with the Snellen chart and converted to the logarithm of minimal angle of resolution (logMAR) units for subsequent analysis, following the standardized guidelines proposed by the Standardization of Uveitis Nomenclature Working Group.^13,14 To explore the effect of early and late treatment on retinal and choroidal microvasculature, patients with VKH disease were divided into two groups according to the time interval between the uveitis onset and the initiation of regular treatment.¹⁵ Group 1 (55 patients) was the early treatment group, starting treatment within 2 months after uveitis onset. Group 2 (48 patients) was the late treatment group, starting treatment 2 months after uveitis onset. Reasons for the late treatment included late referral of patients from other hospitals and patients’ lack of awareness regarding their condition. Eyes with media opacities and other ocular vascular abnormalities were excluded. A total of 103 healthy individuals served as controls. Only the right eye was evaluated in all participants. 

Image Acquisition Protocol

The commercial SS-OCTA device (VG200D; SVision Imaging, Ltd., Luoyang, Henan, China) operates at a wavelength near 1050 nm. It features ultrafast scan speeds of 200,000 A-scans per second, a wide field of view of 56 degrees, and an imaging depth of 2.7 mm in tissue. All the OCTA images covered an area of 6 × 6 mm centered on the fovea using a raster scan protocol of 512 (horizontal) × 512 (vertical) B-scans. The images of retinal and choroidal layers in a 5-mm-diameter concentric annular zone were subdivided into three annular rings with a diameter of 1, 2.5, and 5 mm. The areas with a diameter of 0 to 1 mm and 0 to 5 mm were defined as the central and total area, whereas the areas with a diameter of 1 to 2.5 mm, 1 to 5 mm, and 2.5 to 5 mm were classified into acentric areas. The superficial retinal vascular complex (SVC) spans from an inner boundary located 5 µm above the internal limiting membrane to the lower third of the combined thickness of the ganglion cell layer and inner plexiform layer (GCL + IPL). In contrast, the deep retinal vascular complex (DVC) is defined with its upper boundary at the lower third of the GCL + IPL and its lower boundary 25 µm below the junction between the inner nuclear layer and the outer plexiform layer. The choriocapillaris was 20 µm below Bruch's membrane, and the choroid was automatically identified by VG200D's Van Gogh software. Similar to the methodology employed by the Optovue device (Optovue, Inc., Fremont, California, USA), VD was quantified as the area occupied by blood vessels within a retina projection image.¹⁶ Blood flow area (FA) was utilized as a complementary metric, concentrating on the regions with active blood flow. FA was calculated from angiography projection images by summing the areas of all pixels that surpassed a predefined intensity threshold. The three-dimensional choroidal vascularity index (CVI) of the medium- and large-sized choroidal vessel layer was employed to represent the volumetric vascular density. CVI was defined as the ratio of the luminal choroidal vascular volume (CVV) to the total choroidal volume. 

Upon gathering the OCTA results from all groups, we performed a statistical analysis of the FA and VD metrics in the superficial retinal vessels, deep retinal vessels, and choriocapillaris layer. This was followed by the modeling of three-dimensional choroidal metrics, CVI and CVV. We typically applied VG200D's Van Gogh software and a customized version of a deep learning convolutional neural network called U-Net¹⁷ to identify the contours of the medium- and large-sized choroidal vessels in the B-scans. Subsequently, we reconstructed the three-dimensional morphology of vessels and quantified the medium- and large-sized choroidal vessels, along with their CVI and CVV. The AI algorithms were divided into training and testing sets, and the training set was further augmented by various operations, including brightness and contrast adjustment, scale, shift, and flip, to make the training more scalable to the variation of data collection conditions. A binary cross-entropy loss was then computed and minimized using an Adam optimizer.¹⁸ The network was implemented using Google's TensorFlow framework and trained on a single NVidia (Santa Clara, California, USA) GPU 1070 Ti. To avoid the model trapped in a common statistical error of neural networks called local minima,¹⁹ we repeated the training process for a total of 10 times and selected the one that gave the best evaluation results. Evaluation was done using Accuracy and intersection over union (IoU) from the testing data set. 

Treatment

The patients with VKH disease in this study had all been successfully treated with prednisone combined with immunosuppressive agents according to the protocol described in a previous study²⁰ and had entered an inactive stage of their ocular disease. An initial dosage of 0.6 to 0.8 mg/(kg/d) was used for patients with VKH disease in group 1 and 0.4 to 0.6 mgmg/(kg/d) was used for patients with VKH disease in group 1 and 0.4 to 0.6 mg/(kg/d) for group 2, respectively. This differentiation in dosage was guided by the need to balance efficacy with the potential for adverse effects, considering the severity of the disease in each patient.^2,21 This dosage of corticosteroids was usually used for 1 to 2 weeks in patients with VKH disease and then gradually tapered. A combination with cyclosporine (an initial dosage of 2–4 mg/(kg/d)), cyclophosphamide (an initial dosage of 1–2 mg/(kg/d)), or other immunosuppressive agents, such as azathioprine and methotrexate, was given based on the patient's condition and response to treatment, particularly for those with chronic or recurrent disease who are intolerant or resistant to corticosteroids.¹⁵ Cyclosporine was gradually tapered to a maintenance dosage of 2 mg/(kg/d) over 5 to 6 months if the initial dosage was more than 2 mg/(kg/d), while other immunosuppressive agents were usually used for 3 to 5 months. If the patient developed anterior inflammation during the treatment, corticosteroid eye drops and cycloplegic agents were also used. For patients with relapsing uveitis, a modified regimen was adopted. This included oral prednisolone, azathioprine, and cyclosporine, which has been reported to result in rapid remission and help prevent recurrences.²² In such cases, the initial dose of corticosteroids and immunosuppressive agents was reused, and corticosteroid and cycloplegic eye drops were added. The entire treatment usually lasted more than 1 year, and by the time we started the current study, 45 patients with VKH disease had stopped medication. 

Data Analysis and Statistical Methods

For statistical analysis, we used the right eye from each patient. Categorical variables of basic information are shown as numbers and percentages, whereas continuous variables are shown as the mean ± standard deviation (SD) along with the 95% confidence interval for the mean and were analyzed with SPSS software (version 23.0; SPSS, Inc., Chicago, IL, USA). The differences between gender within groups were determined by the χ² test. One-way analysis of variance was used to test for differences among patients with VKH disease and normal controls, and post hoc tests were used between group pairs. The independent t-test and Mann–Whitney U test were used for the comparison between group 1 and group 2, controls and patients with VKH disease, respectively. Wilcoxon matched-pairs signed-ranks test was performed to evaluate the change of BCVA between the first and last visits. Correlation analysis was applied by GraphPad software version 6.02 (GraphPad Software, San Diego, CA, USA) to evaluate the relationship between OCTA parameters and final logMAR BCVA. A P value less than 0.05 was considered statistically significant. 

Results

Patient Characteristics

A total of 103 patients with VKH disease and 103 normal controls were enrolled in the study. There were no significant differences between the normal controls and groups of patients with VKH disease concerning age, gender, IOP, and duration of VKH disease (P = 0.261) (Table 1). Compared to the initial BCVA, visual outcomes had improved at the last follow-up in 53 (96.4%) patients of group 1 (P < 0.001) and 44 (91.7%) patients of group 2 (P = 0.006), respectively. Compared to group 1, the initial logMAR BCVA was better in group 2 (group 1 was 0.44 ± 0.46, group 2 was 0.42 ± 0.74, P = 0.012), and the final logMAR BCVA showed no significant difference (group 1 was 0.11 ± 0.17, group 2 was 0.15 ± 0.24, P = 0.225). In all patients with VKH disease, the initial BCVA before treatment was more than 0.1 logMAR (<20/25 on the Snellen chart) in 68 (66.0%) eyes, and the average logMAR value was 0.4 ± 0.6. The final follow-up BCVA was significantly improved (P < 0.001), whereby 31 (30.1%) eyes had a logMAR value of more than 0.1 (<20/25 on the Snellen chart), and the average logMAR value was 0.1 ± 0.2 (Table 1). 

Table 1.

 

Characteristics of Patients With Vogt–Koyanagi–Harada (VKH) Disease and Normal Controls

 

Retinal FA and VD Measurements

The FA and VD of the retinal SVC and DVC in the five annular zones (the total area, diameter of 0–1 mm, 1–2.5 mm, 1–5 mm, and 2.5–5 mm areas) around the fovea were compared between the VKH disease groups and healthy controls. The retinal OCTA values are shown in Table 2, and typical examples of a normal control (Figs. 1A–D) and patients with VKH disease from the two groups (Figs. 1E–L) are shown. 

Table 2.

 

Retinal SS-OCTA Measurements in Patients With VKH Disease and Normal Controls

 

Figure 1.

 

Representative retinal SS-OCTA and postprocessed images for a control eye and inactive Vogt–Koyanagi–Harada (VKH) eyes. The images were acquired in 6 × 6-mm areas around the fovea. (A–D) OCTA examinations of a control eye. (E–H) OCTA examinations of a patient with inactive VKH who received early treatment <2 months after the onset of uveitis. (I–L) OCTA examinations of a patient with inactive VKH who received late treatment >2 months after the onset of uveitis. (A, E, I) FA in the SVC of a control eye, an inactive VKH eye with early treatment, and an inactive VKH eye with later treatment, respectively. The brighter color indicates the larger FA value. (B, F, J) VD in the SVC of a control eye, an inactive VKH eye with early treatment, and an inactive VKH eye after late treatment, respectively. The warmer color indicates the larger VD value. (C, G, K) FA in the DVC. (D, H, L) VD in the DVC.

In the SVC, FA and VD in group 2 were significantly lower than in the healthy control group, while these parameters in group 1 were not different from the controls. Specifically, the FA of group 2 in total and acentric areas had decreased values (total area, 1–2.5 mm, 1–5 mm, 2.5–5 mm, all P < 0.001), and only the acentric areas showed a decreased VD (1–2.5 mm P = 0.001). When comparing the two VKH groups, group 2 showed a significantly lower FA than group 1 (total area P < 0.001, 1–2.5 mm P = 0.006, 1–5 mm P < 0.001, 2.5–5 mm P < 0.001, respectively). There was no significant difference in VD of the SVC between the two VKH disease groups. 

In the DVC, the FA and VD of the two VKH disease groups were all statistically lower than that found in the healthy control eyes. In group 1, the total and acentric FA (total area P = 0.001, 1–2.5 mm P = 0.001, 1–5 mm P < 0.001, 2.5–5 mm P < 0.001, respectively) and acentric VD (1–5 mm P = 0.003, 2.5–5 mm P = 0.004, respectively) had decreased values when compared to controls. In group 2, the FAs of total, central, and acentric areas were all lower than the control eyes (total area P < 0.001, 0–1 mm P = 0.001, 1–2.5 mm P < 0.001, 1–5 mm P < 0.001, 2.5–5 mm P < 0.001, respectively) and the acentric VD had decreased values (1–2.5 mm P = 0.042, 1–5 mm P = 0.003, 2.5–5 mm P < 0.001, respectively). When comparing the two patient groups, group 2 showed a significantly lower FA than group 1 (total area P < 0.001, 0–1 mm P = 0.003, 1–2.5 mm P = 0.012, 1–5 mm P < 0.001, 2.5–5 mm P < 0.001, respectively). There was no significant difference in VD of the DVC between the two VKH disease groups. 

FA of the Choriocapillaris, CVV, and CVI of Medium- and Large-Sized Choroidal Vessel Layer Measurements

The FA values of the choriocapillaris were all significantly decreased in the two VKH disease groups when compared to healthy controls, and flow void areas could be identified in the VKH disease eyes. In group 1, the total, central, and acentric FAs had decreased values (total area P = 0.007, 0–1 mm P < 0.001, 1–2.5 mm P = 0.007, respectively). In group 2, the FAs of all areas showed deceased values (total area P < 0.001, 0–1 mm P < 0.001, 1–2.5 mm P < 0.001, 1–5 mm P = 0.004, 2.5–5 mm P = 0.008, respectively). When comparing the two patient groups, group 2 showed a significantly lower acentric FA than group 1 (1–2.5 mm P = 0.033) (Table 3). 

Table 3.

 

Choroidal SS-OCTA Measurements in Patients With VKH Disease and Normal Controls

 

CVV of medium- and large-sized choroidal vessel layer was obviously decreased in group 2 when compared to controls, while group 1 showed relatively higher CVV but no significant difference when compared to controls (all P > 0.05). In group 2, the CVV of total and acentric areas was lower than the controls (total area P = 0.049, 1–5 mm P = 0.045, 2.5–5 mm P = 0.041, respectively). Moreover, the CVV of group 2 was significantly lower than that in group 1 in all annular zones (total area P = 0.011, 0–1 mm P = 0.037, 1–2.5 mm P = 0.017, 1–5 mm P = 0.011, 2.5–5 mm P = 0.012, respectively). There was no significant difference in CVI when comparing group 1, group 2, and normal controls with each other (all P > 0.05) (Table 3). 

Typical examples of a normal control (Figs. 2A–C) and representative patients with VKH disease from the two groups (Figs. 2D–I) are shown. 

Figure 2.

 

Representative choroidal SS-OCTA and postprocessed images for a control eye and inactive VKH eyes. The images were acquired in 6 × 6-mm areas around the fovea. (A–C) OCTA examinations of a control eye. (D–F) OCTA examinations of a patient with inactive VKH who received early treatment <2 months after the onset of uveitis. (G–I) OCTA examinations of patients with inactive VKH disease who received later treatment >2 months after the onset of uveitis. (A, D, G) FA in the choriocapillaris of a control eye, an inactive VKH eye with early treatment, and an inactive VKH eye with late treatment, respectively. The brighter color indicates the larger FA value. FA images showed choroidal capillary hypoperfusion areas (red arrows). (B, E, H) The three-dimensional CVV recognition in medium- and large-sized choroidal vessel layers. The original shape of the choroidal vessel is shown on the right, and the map of choroidal vessels can be seen in the bottom left box. The warmer the color, the higher the CVV. (C, F, I) The three-dimensional CVI recognition in medium- and large-sized choroidal vessel layers. The choroidal vessels were identified and marked yellow in the right box. The map of choroidal vessels can be seen in the bottom left box. The warmer the color, the higher the CVI.

The Correlation Between OCTA Parameters and the Final BCVA

The final logMAR BCVA at the last follow-up visit to our uveitis clinic was significantly improved in both VKH disease groups when compared to the first visit. To investigate the total (area of 0–5 mm in diameter), central (area of 0–1 mm in diameter), and acentric (area of 1–5 mm in diameter) OCTA values on visual outcome, we performed a correlation analysis (Fig. 3). The final logMAR BCVA was significantly negatively correlated with the FA of retinal SVC (total area P = 0.0002, central area P < 0.0001, acentric area P = 0.0011, respectively), FA of retinal DVC (P < 0.0001 in all areas), VD of retinal DVC (acentric area P = 0.0017), FA of the choriocapillaris (total area P = 0.0005, central area P < 0.0001, acentric area P = 0.0047, respectively), and CVI of medium- and large-sized choroidal vessels (acentric area P = 0.0282). However, VD of retinal SVC or CVV of medium- and large-sized choroidal vessels did not correlate with any area of the quantitative parameters of OCTA (all P > 0.05). 

Figure 3.

 

The correlation analysis of BCVA and SS-OCTA parameters in the total area, central area, and acentric area.

Discussion

In this study, we compared retinal and three-dimensional changes of the choroidal microvasculature in VKH disease eyes from patients who received an early or late treatment using quantitative OCTA metrics of FA, VD, three-dimensional CVV, and CVI. The study shows that early treatment of VKH disease results in a better retinal and choroidal microvascular outcome. Accompanying software allows AI-assisted measurements of the three-dimensional CVV and CVI to meet growing clinical needs in exploring the deep choroidal layers. To the best of our knowledge, the effect of treatment delay in VKH disease on the microvasculature of the retina and choroid has not yet been reported, and our study provides further evidence that patients with VKH disease should be treated as soon as possible. 

In the present OCTA study, the metrics of the retinal microvasculature showed that in the retinal SVC, the FA and VD in inactive patients with VKH disease who had received a delayed treatment were statistically lower than values observed in healthy control eyes. This was obvious for the total as well as for the acentric regions of the macula. Distinctly, the FA and VD of patients treated within 2 months after uveitis onset were not different from the controls. Similar findings regarding the SVC across different groups have been reported in other studies. For instance, patients with sunset glow fundus (SGF) exhibit significantly lower VD in the superficial retina compared to both normal controls and patients without SGF.¹⁰ Additionally, patients with VKH disease with anterior uveitis recurrence demonstrate significantly lower VD compared to convalescent patients.²³ The possible explanation for this might be that the inflammation starts in the choroid and deep retina and that the superficial retina is affected later in the disease process. As mentioned earlier, the microvascular impairment is most likely a result of local inflammation.^24,25 When the treatment was timely and effective, the inflammation might be limited to the deeper retina adjacent to the choroid and will not yet affect the superficial retina. Early treatment may prevent extensive inflammatory damage by modulating the immune response and reducing the proinflammatory cytokine release that leads to vascular damage. Delayed treatment, on the other hand, allows prolonged inflammation, resulting in persistent vascular endothelial damage, increased vascular permeability, and subsequent ischemia and hypoxia of the retinal and choroidal tissues. Similar to other tissues with ample blood supply and circulation, ischemia in the choroid and retina can lead to further upregulation of inflammatory mediators and perpetuate a cycle of vascular injury.²⁶ 

Our results showed a decreased FA and VD of DVC in all patients with VKH disease, whether they were treated early or late. Our results are in agreement with earlier studies showing that the metrics of DVC are sensitive parameters for monitoring disease status and treatment response.^8,23 The damaged transmission function²⁷ and insufficient supply of aerobic glycolysis²⁸ might also play an important role and may explain the observed DVC abnormalities. The acentric metrics of the retina were all statistically decreased in patients with VKH disease, although the central area (diameter of 0–1 mm) only showed a decreased FA of the DVC in patients receiving late treatment. The central area includes the foveal avascular zone (FAZ), and some studies have reported that the FAZ area increased significantly in quiescent patients with VKH disease, which would certainly affect the measurements in these regions.²⁹ More studies need to be done to explore the OCTA results after excluding the FAZ from each region. In addition, we found significant differences between group 1 and group 2 in FA of the SVC and DVC, indicating that the treatment within 2 months after the onset of the disease was of great significance for the recovery in retinal flow and microvasculature. We described an association with more extensive vascular damage to the superficial retina only in patients with VKH disease with delayed treatment, while the deep retina was affected in all patients with VKH disease. 

In this study, we also evaluated the OCTA analysis of the choriocapillaris. Overall, FA of the choriocapillaris was significantly lower in both VKH groups when compared to healthy controls, and multifocal flow void spots in the choriocapillaris could be observed clearly. Several studies have used flow void analysis to reveal damage of the choroid in inactive VKH disease.^9,30,31 However, intercapillary spaces are a natural aspect of the normal choroidal structure. Additionally, signal loss foci may indicate the physiologic enlargement of vascular spaces, influenced by factors such as aging, hypertension, and myopia, which can reduce the overall average detected blood flow in the choriocapillaris layer.³² In our study, we therefore used FA instead of flow void and obtained similar results concerning damage of the choriocapillaris, as reported earlier by others.^9,30,31 An earlier study observed a normal FA and flow density in the choriocapillaris of patients with inactive VKH disease, which is not in agreement with our findings.²⁹ The discrepancy might be due to the small sample size and the different stratification of the choriocapillaris. The decreased FA of the choriocapillaris suggests an impairment in the choriocapillaris microvasculature, and the lower parafoveal value in the late treatment group shows that timely treatment may facilitate the restoration of the choriocapillaris. 

Our study differs from earlier studies in that we used three-dimensional CVV to investigate medium- and large-sized choroidal vessels. Several studies defined medium- and large-sized choroidal vessel layers by the distance between different interfaces, which could make the measured values less accurate and included a large amount of choroidal stroma besides blood vessels.^33,34 Therefore, the choroidal microvasculature should be accurately contoured and measured using three-dimensional metrics. To our knowledge, we are the first to report that late treatment of VKH disease is associated with a significant decrease in the AI-assisted three-dimensional CVV. We hypothesize that the local autoimmune response against melanin-associated antigens causes a local choroidal inflammation, which subsequently results in an impaired blood flow in the deep choroid. Early choroidal ischemia and hypoxia could cause vascular contraction, damage of the vascular endothelium, and a thickened basement membrane, which reduces choroidal blood flow filling that subsequently leads to choroidal vascular stenosis, regression, and even atrophy. The decrease in choroidal volume might further promote tissue hypoxia and increase vascular endothelial growth factor levels.³⁵ Furthermore, we hypothesize that early intervention mitigates the inflammatory cascade more effectively, preserving the structural and functional integrity of the choroidal vasculature. In contrast, delayed treatment likely allows for the chronicity of inflammation, leading to irreversible changes and eventual destruction of the vascular network. These pathologic changes highlight the critical window during which immunomodulatory therapies can be most beneficial. Early treatment in VKH disease is beneficial, as evidenced by the fact that there was no significant difference between group 1 and controls concerning the CVV. The CVV of group 1 was even slightly larger than that seen in controls, suggesting that the early treatment might also promote telangiectasis and repair of mild- and large-sized choroidal vessels in order to support the oxygen supply and nutritional demand. 

To the best of our knowledge, this was also the first study to investigate three-dimensional CVI in patients with inactive VKH disease and showed no significant differences in CVI among the patients and controls. This result suggests that three-dimensional CVI could be used to define the general inactive VKH disease status regardless of whether patients with VKH disease were treated early or late. As we found that CVI did not significantly change whereas CVV was decreased in patients who had undergone a delayed treatment, we hypothesized that the total choroidal volume might be smaller in patients treated late as compared to patients who had been treated early and normal controls. Others showed a higher two-dimensional CVI in acute phase VKH disease, which returned to normal values in patients with inactive VKH disease, which is consistent with our results.³⁶ If VKH disease becomes chronic, this is associated with a prolonged two-dimensional CVI increase.³⁷ The increased CVI might be due to the limited two-dimensional measuring region of 1500 µm and active metabolism of the sunset glow fundus.³⁸ Since some reports showed that CVI could be variable in different regions, we validated the three-dimensional metrics by using Accuracy and IoU software using a testing data set.³⁹ The coefficient of Accuracy was 0.97, and IoU was 0.75, suggesting a good reproducibility. 

This study showed that metrics of FA were more useful in showing an association with visual outcome than the VD, CVI, and CVV measurements. In particular, FAs of SVC, DVC, and the choriocapillaris were all significantly associated with the final BCVA, while VD of SVC and CVV of the medium- and large-sized choroidal vessels did not reveal an association with a final visual outcome. The explanation for this finding may be due to the fact that the microvascular blood flow plays a more important role in BCVA. Based on these findings, more attention should be paid to the blood flow changes rather than focusing on vessel density. A practical consequence is that patients with VKH disease should undergo auxiliary examinations related to fundus flow when their visual acuity decreases. 

Taken together, this study shows that three-dimensional OCTA of the chorioretinal microvasculature provides further evidence that patients with VKH disease should be treated as soon as possible. 

Conclusions

In summary, our investigation revealed that early treatment of VKH disease within 2 months of onset results in significantly better outcomes for chorioretinal microvasculature, as evidenced by OCTA metrics, compared to late treatment. Early treated patients demonstrated higher flow area and vessel density in both the superficial and deep vascular complexes, as well as in the choriocapillaris and choroidal vascular volume, relative to those treated later and to healthy controls. Notably, no significant difference was observed in the choroidal vascularity index among groups. Furthermore, parameters such as flow area and vessel density were inversely correlated with the final logMAR BCVA, highlighting their potential as predictive markers for visual outcomes. This underscores the importance of timely intervention in VKH disease to preserve chorioretinal microvascular integrity and optimize visual prognosis. 

Acknowledgments

The authors thank Junchao Peng, Qingfeng Wang, the First Affiliated Hospital of Chongqing Medical University, and Chongqing Eye Institute, Chongqing, P.R. China, for their technical help in this study. The authors thank all the patients and healthy controls who participated in this study. 

Supported by Key Project of Chongqing Science and Technology Bureau (CSTC2021jscx-gksb-N0010), Chongqing Outstanding Scientists Project (2019), Chongqing Chief Medical Scientist Project (2018), Chongqing Key Laboratory of Ophthalmology (CSTC, 2008CA5003), and Chongqing Science & Technology Platform and Base Construction Program (cstc2014pt-sy10002). 

Disclosure: F. Huang, None; S. Tan, None; J. Hu, None; R. Hu, None; P. Yang, None