Subretinal fluid disturbs the retinal venous blood flow in central serous chorioretinopathy

The significance of subretinal fluid in the retinal blood flow is unclear. Here, we evaluated the association between subretinal fluid (SRF) and retinal blood flow in eyes with central serous chorioretinopathy (CSC) using a retinal functional imager (RFI) and optical coherence tomography angiography (OCTA). In this retrospective case–control study involving 26 eyes from 18 CSC patients and 25 eyes from 21 age- and sex-matched controls, we found that the CSC group showed significant differences from the control group in terms of the retinal venule blood flow velocity (3.60 ± 0.43 vs 3.96 ± 0.56 mm/s; p = 0.030), retinal venule blood flow rate (8.75 ± 2.67 vs 12.51 ± 7.12 nl/s; p = 0.040), and the diameter of retinal venules (118.26 ± 14.25 vs 126.92 ± 35.31 μm; p = 0.045). Linear regression analysis showed that SRF thickness accounted for a 36.9% reduction in venous BFR (p = 0.013). The difference in the O2 saturation between retinal arteries and veins was greater in the CSC group. There was no correlation between SRF thickness and capillary densities in OCTA. Our findings suggest that disturbance in venous return and the associated altered oxygen may be significant changes in the retinal blood flow dynamics in eyes with SRF.


Relationship between SRF thickness and OCTA parameters.
Retinal capillary densities (FCP, SCP, and DCP) decreased as the SRF thickness increased, albeit without a statistically significant correlation ( Fig. 2ac). The FAZ area was also not significantly associated with the SRF thickness (Fig. 2d).

Representative cases. Case 1.
A 47-year-old male presented with blurred vision on his right eye for 6 months. On examination, the visual acuity was 20/20 in the left eye. The anterior segments of both eyes were found to be normal by slit-lamp biomicroscopy. Fundus examination of the left eye was normal (Fig. 3a). OCT revealed normal in the left eye (Fig. 3b). O 2 saturation image was taken by RFI (Fig. 3c). In red are the arteries that do not lose oxygen. Small venules near the fovea still have high oxygen saturation because they are very close to the arterial capillary. Blood of small venules drained to the superior and inferior retinal arcade veins. The O 2 saturation of superior and inferior retinal arcade veins reached 25-75%. The number of retinal vein segments of blue color was lesser than that of the retinal vein segments of green-to-red color (Fig. 3c). Arterial BFV of the left eye was −4.674 mm/s and venous BFV of the same eye was 4.501 mm/s (Fig. 3d).

Case 2.
A 50-year-old male presented with blurred vision on his left eye for 2 years. On examination, the visual acuity was 20/30 in the left eye. The anterior segments of both eyes were found to be normal by slit-lamp biomicroscopy. Fundus examination and OCT of the left eye showed well-demarcated SRF on macular (Fig. 4a,b). O 2 saturation image was taken by RFI (Fig. 4c). The color of O 2 saturation of retinal arteries was darker than that of Case 1. The O 2 saturation of retinal arteries near the fovea was greater than that of Case 1 (Figs. 2c, 3c). Small venules near the fovea and crossing arteries still had high oxygen saturation (Fig. 4c). The O 2 saturation of superior and inferior retinal arcade veins reached a similar level with that of Case 1, but the number of retinal vein segments of blue color was greater than that of the retinal vein segments of green-to-red color (Fig. 4c). Arterial BFV of the left eye was −3.991 mm/s and venous BFV of the same eye was 3.676 mm/s (Fig. 4d).

Discussion
CSC is characterized by delayed choroidal infusion, choroidal vascular hyperpermeability, and choroidal venous dilation, which suggests that the choroid is primarily involved in the pathophysiology of CSC 20,21 . However, with the development of OCTA, recent studies have reported that CSC cases have blood flow abnormalities in the external retina 22 . As an example, Nelis et al. reported an increase in retinal flow density and a decrease in the FAZ in both the affected and unaffected eyes of CSC patients 9 . Therefore, retinal vascular problems may also be involved in the pathophysiology of CSC.  www.nature.com/scientificreports/ Several imaging techniques are available for retinal vascular analysis, including FA, OCTA, adaptive optics with scanning light ophthalmoscopy, and laser speckle flow cytometry. Compared with these imaging techniques, RFI has advantages in quantitative measurement of the blood flow velocity and oxygen saturation [23][24][25][26] . Moreover, RFI is non-invasive and does not require intravenous fluorescein dye injection 18,26 . However, RFI has a disadvantage compared with OCTA in terms of 3D imaging and volumetric imaging 17,27 . In the present study, we performed the measurement of BFV and BFR using RFI in eyes with or without CSC. As expected, venous BFR and BFV in the fovea were significantly smaller in the CSC group than in the control group. Arterial BFV was positively correlated with BFR and negatively correlated with arterial vessel diameter. Venous BFV was positively correlated with BFR and venous vessel diameter. DCP density was positively correlated with venous BFV, and SRF thickness was negatively correlated with venous BFR.
Beutelspacher et al. observed that the retinal BFV in the retinal veins was significantly smaller in CSC, especially in the larger retinal veins, while retinal arterial BFV was not significantly different between CSC patients and controls 5 . The result of the present study is in line with the previous study as we found that the diameters of retinal veins were smaller in CSC, while those of retinal arteries were not significantly different from the control group (Table 1). We assume that the lack of significant differences in BFV, BFR, and vessel diameter in retinal arteries in CSC is due to the presence of autoregulation in retinal arteries [28][29][30] . In contrast, the reason for the significant decreases in BFV, BFR, and vessel diameter of retinal veins in CSC is likely due to the mass effect of  showed that CSC patients had higher DCP density and larger FAZ area than did healthy controls 9 . In our study, CSC patients had a lower FCP density, smaller FAZ area, and higher DCP density than did healthy controls, albeit without statistical significance ( Table 1). The lack of significant difference in the DCP density and FAZ area according to the presence of CSC may be due to the small sample size. Nelis et al. suggested that the decrease in FAZ and increase in DCP density in CSC may be due to pathological changes in CSC such as accumulation of SRF 9 , and Eperon et al. suggested that continuous SRF may stimulate an increase in retinal flow density 31 . There was no correlation between SRF thickness and capillary densities in our study. www.nature.com/scientificreports/ In our study, arterial BFV in CSC patients was negatively correlated with the arterial vessel diameter, especially in large retinal arteries, which may be the result of the autoregulation of retinal arteries. If the BFR of retinal arteries is increased, the autoregulation system may suppress the input of blood flow by decreasing the retinal arterial vessel diameter in order to maintain the BFR of retinal capillaries at a certain level. Also, venous BFV was positively correlated with venous BFR, venous vessel diameter, and DCP density. The DCP was organized into capillary vortexes (i.e., radial convergence of capillaries toward an epicenter), which drain into the superficial venules 32 . Therefore, as the vascular DCP increases, the BFR and BFV of retinal veins may increase because of the increased flow from the DCP.
The difference in the O 2 saturation between retinal arteries and veins was greater in CSC patients than in controls. Li et al. used a non-invasive retinal oximeter (Oxymap T1, Oxymap ehf., Reykjavik, Iceland) to show that in the eyes of CSC patients, the O 2 saturation of retinal arteries was increased in the inferotemporal quadrant and the O 2 saturation of retinal veins was decreased in the inferonasal quadrant 33 . In addition, O 2 saturation of retinal arteries also increased around the macular region, suggesting that these contribute to the pathophysiology of CSC 33 . Our study showed similar results with Li et al. 's study. Turkcu et al. found that the antioxidant capacity was significantly decreased in CSC cases, suggesting that the antioxidant defense system may be inadequate or corrupted in CSC 34 . Moreover, lower retinal vein velocity leads to more residence time of blood flow, which may further drive the difference in O 2 saturation between retinal arteries and veins. It is also possible that the surrounding SRF may dilute the retinal venous O 2 saturation.
Our study has several limitations. First, this study could have been underpowered to detect small differences according to the presence of CSC because of the small sample size. Most of the eyes had chronic CSC, and www.nature.com/scientificreports/ parameters such as blood pressure and heart rate were not measured in this study. As recruiting larger cohorts of CSC patients may be a challenge in single centers, multicenter studies on this issue are warranted. Nevertheless, this study provides the information of correlation between BFV, BFR, and the retinal capillary density and SRF thickness in CSC eyes. Second, RFI is non-invasive but relatively time-consuming (2-10 min) and thus requires patience in the examinees. Lastly, because CSC predominantly affects people aged between 20 and 50 years, there could have been a selection bias in the patient group; nevertheless, we tried to overcome this issue by matching the control group by age.
In conclusion, we found that SRF in eyes with CSC affected the retinal homeostasis through alterations in blood flow. Venous BFV, BFR, and vessel diameter were lower in eyes with CSC than healthy eyes. The SRF thickness was associated with reduced venous BFR, and CSC may aggravate the SRF due to venous stasis. Compared with OCTA, RFI had advantages in detecting subtle changes in retinal blood flow in the presence of SRF. The difference in O 2 saturation between retinal arteries and veins was greater in CSC than in controls. These findings suggest that retinal venous problems and the associated altered oxygen metabolism may be significant changes in the retinal blood flow dynamics in eyes with SRF.

Methods
Subjects. This retrospective study included patients with CSC and age-and sex-matched controls. The data were collected between October 2019 and February 2020 at the Department of Ophthalmology in Yeungnam University Hospital (Daegu, South Korea). Of the patient data, we selected those from patients who had undergone OCT, OCTA, and RFI. All patients had undergone comprehensive ophthalmic examinations including www.nature.com/scientificreports/ best-corrected visual acuity (BCVA) measurement, dilated fundus examination, spectral-domain OCT (SD-OCT, Spectralis; Heidelberg Engineering, Heidelberg, Germany), OCTA (Optovue, Inc, Fremont CA, USA), and RFI 3005 (Optical Imaging Ltd., Rehovot, Israel). This retrospective study was approved by the institutional review board of Yeungnam University Hospital and adhered to the tenets of the Declaration of Helsinki for research involving human subjects. Informed consent was obtained from all subjects. All the methods were performed in accordance with relevant guidelines/regulations. The clinical diagnosis of CSC was based on the presence of subretinal fluid (SRF) with or without pigment epithelium detachment (PED) accompanied by local and/or diffuse leakage in FFA and ICGA. The exclusion criteria were as follows: (1) retinal diseases that affect ocular circulation such as diabetic retinopathy and retinal vein occlusion, (2) concomitant retinal disease such as retinal detachment, macular hole, epiretinal membrane, and glaucoma, (3) refractory error more than 3.5 diopters, (4) massive subretinal hemorrhage or fibrosis obscuring the choroidal vasculature on RFI and OCTA, (5) severe media opacity that could degrade image quality, and (6) history of treatments that can cause significant changes to choroidal status such as intraocular surgery. Cataract surgery performed more than 3 months previously was not considered an exclusion criterion. Agematched patients without ocular disease as confirmed by history and ophthalmic examinations who visited our clinic for a health-screening checkup were included as healthy controls. We randomly selected one of the two eyes in the healthy control group.
Retinal microvascular blood flow and O 2 saturation analysis using RFI. Flowmetry. The RFI system (Optical Imaging, Rehovot, Israel) is composed of a fundus camera, stroboscopic illumination, and a fast filter wheel 17 . Fast stroboscopic illumination enables the camera to take multiple snapshots of the retina in less than 0.2 s to track the movement of red blood cells through each sequential frame 16,18 . Using multiple sequences, RFI creates capillary perfusion maps and semi-automatically assesses blood flow velocity in arterioles and venules 16 . The instrument's analysis software provides quantitative analysis of the BFV that includes the following: BFV in each segmental artery, vein, and total, identification number of each segment that was marked, and the diameter of each blood vessel.
To assess the regional BFV and BFR, a grid with a 1.5-mm diameter ring centered on the fovea was applied to define two zones-fovea (interior area of the 1.5-mm diameter ring) and parafovea (exterior area of the 1.5mm diameter ring).
Oximetry. Qualitative blood oximetry maps were generated from different reflections of the retinal vasculature using varying wavelengths according to previously described protocols 16 .
Central macular thickness and subretinal fluid thickness measurement. CMT and SRF thickness was measured in the subfoveal region using OCT. CMT was defined as the vertical distance between the internal limiting membrane and the top of the RPE at the fovea. CMT was quantitatively analyzed by a custom-built software program (Heidelberg Eye Explorer, Version 1.9.10, Heidelberg Engineering, Heidelberg, Germany). SRF thickness was defined as the greatest distance between the top of the SRF and the top of the RPE at the fovea. Measurement of SRF thickness was performed using the built-in caliper tool in the Heidelberg Eye Explorer program at a single point below the fovea by an independent masked grader (C.M.). The supervising grader (J.L.) confirmed the final decision.
Measurement of vessel density and area of foveal avascular zone. OCTA (Optovue, Inc, Fremont, CA, USA) was performed using the split spectrum amplitude-decorrelation angiography algorithm. For each eye, 3 × 3 mm-sized images were taken and analyzed at the full capillary plexus (FCP) and two capillary layers-superficial capillary plexus (SCP) and deep capillary plexus (DCP). FCP was defined as the retinal layer between the internal limiting membrane and the RPE. SCP was imaged with an en face section starting at the inner border of the ganglion cell layer to the inner border of the inner plexiform layer in the macular area. En face images of the DCP were obtained by segmenting from the inner boundary of the inner plexiform layer to the outer boundary of the outer plexiform layer 19 .
The segmentation was identified automatically. Vascular density was calculated as the percentage of the area occupied by vessels in the total area of images and the selected depth of vessels. Vascular density was calculated in 3 × 3 mm-sized images in the SCP, DCP, and FCP. FAZ area was defined as the area of capillary-free region demarcated by a ring of interconnecting capillaries at the margin of the fovea. Measurement of the SCP density, DCP density, FCP density, and FAZ area was performed using en face imaging and EnView data analysis software (AngioAnalytics, Optovue).

Statistical analysis.
Statistical analyses were performed using IBM SPSS V.20.0 for Windows (IBM Co., Armonk, NY, USA). The Mann-Whitney test and Chi-squared test were used for comparison of numerical variables between CSC patients and age-matched healthy controls. Pearson's correlation test was performed between numerical variables and BFV of each size of vessels. Linear regression analysis was performed between SRF thickness and arteriole and venule BFV and BFR. Variables with a p value < 0.05 were considered statistically significant. All results are presented as mean ± standard deviation.

Data availability
All data generated or analyzed during this study are included in this published article. Additional datasets or raw files during the current study are available from the corresponding author on reasonable request.