Evaluation of flow of chorioretinal capillaries in healthy black and white subjects using optical coherence tomography angiography

This study compared macular capillary parameters between healthy black and white subjects using optical coherence tomography angiography (OCTA). We measured vessel density (VD) of superficial (SCP), intermediate (ICP), and deep (DCP) capillary plexuses and choriocapillaris blood flow area (BFA) of the fovea, parafovea and total 3 mm-diameter circular area centered on the fovea, as well as the foveal avascular zone (FAZ) parameters, controlling for axial length. Black subjects had lower foveal and parafoveal VD in the SCP (p = 0.043 and p = 0.014) and the ICP (p = 0.014 and p = 0.002). In the DCP, black subjects had a trend toward lower foveal and parafoveal VD. Black subjects had decreased choriocapillaris BFA in the total 3 mm area (p = 0.011) and the parafovea (p = 0.033), larger FAZ area (p = 0.006) and perimeter (p = 0.014), and a higher capillary density in a 300 μm wide region around the FAZ (FD-300) (p = 0.001). There was no significant difference in FAZ acircularity index. To our knowledge, this is the first report analyzing the three distinct retinal capillary plexuses and identifying differing baseline VD, choriocapillaris and FAZ parameters in healthy young black compared to white subjects. Larger studies are needed to validate these findings and better understand racial differences in vulnerability to ocular diseases.

Morphological studies in humans and animal models have demonstrated the presence of three distinct capillary plexuses in the macula: superficial capillary plexus (SCP), intermediate capillary plexus (ICP), and deep capillary plexus (DCP). It was shown that the inner plexiform and the outer plexiform layers of the retina, supplied by the ICP and the DCP respectively, are particularly vulnerable to ischemic injury due to these layers' high metabolic activity [1][2][3] . The visualization of ICP in particular was difficult in the past because of this layer's thinness, and imaging artifacts 4,5 . Optical coherence tomography angiography (OCTA), rapid and non-invasive imaging modality optimal for assessing retinal and choroidal vascular parameters [6][7][8] , has allowed for the first time to visualize and analyze the three distinct retinal capillary plexuses in vivo with the help of novel projection artifact removal (PAR) software and improved segmentation 9,10 . The ability to evaluate the microvasculature of all three capillary layers in healthy and diseased eyes may significantly enrich our understanding of retinal disorders.
Diabetic retinopathy and retinal vein occlusion, associated with ischemic retinal injury, have a more aggressive presentation in black populations [11][12][13] . Glaucoma, recently shown to be associated with decreased retinal capillary perfusion within the superficial capillary plexus 14 , is known to be more prevalent and more severe in black as compared to white people [15][16][17][18][19] . Better understanding of baseline characteristics of the three retinal capillary plexuses and the choriocapillaris in black as compared to white subjects may contribute to our understanding of the role of race as a risk factor in retinal vascular diseases and glaucoma. Our study aims at comparing baseline retinal vascular parameters including vessel density (VD) of SCP, ICP and DCP, choriocapillaris blood flow area Analyzing three retinal capillary plexuses and the FAZ. We used customized segmentation for the SCP, ICP, and DCP in our analysis, as described by Nesper et al. 21 .The SCP boundaries were defined from the internal limiting membrane (ILM) to 10 um above the inner plexiform layer (IPL) to encompass the nerve fiber and ganglion cell layers. The ICP was defined from 10 um above to 30 um below the IPL to encompass the IPL. The DCP was defined from 30 um below the IPL to 10 um below the outer plexiform layer (OPL) to encompass the OPL (Fig. 1).
Vessel density measurements and FAZ parameters. Using the AngioAnalytics software, we calculated VD as the percentage of the image occupied by retinal blood vessels within the SCP, ICP, and DCP layers in the fovea, parafovea and a 3-mm-diameter circle centered on the fovea. The foveal area was defined as a 1-mmdiameter circle centered on the fovea. The parafoveal area was defined as a ring from the edge of the foveal circle to the edge of a 3-mm-diameter circle centered on the fovea (Fig. 2). The FAZ size (in mm 2 ) was automatically calculated by the AngioAnalytics software using the nonflow area measurement tool. FAZ perimeter (mm) was defined as the length of the contour demarcating the FAZ. FAZ acircularity index (AI) was calculated as the ratio of the FAZ perimeter to the perimeter of a circle with an equal area, AI = FAZ Perimeter/√(4 × π × Area) 13 . We also recorded FD-300, a parameter calculated by the AngioAnalytics software as retinal capillary density from ILM to OPL in a 300 μm wide region around the FAZ (Fig. 3).
Analysis of the choriocapillaris blood flow area. Blood flow area (BFA), the percentage of a retinal image occupied by pixels corresponding to blood vessels, was calculated automatically after the image capture process by AngioAnalytics software 15 . In this study we obtained choriocapillaris BFA within 1 mm-and 3 mmdiameter circular zones centered manually on the fovea. The choriocapillaris BFA in the parafoveal region was determined by subtracting the BFA of the 1 mm-diameter circular area from the BFA of the 3 mm-diameter circular area centered on the fovea (Fig. 2 www.nature.com/scientificreports/ evaluate the association between race and mean macular capillary parameters, a mixed-effects linear regression model was used, accounting for 2 eyes of the same subject, and adjusting for axial length. A p-value of < 0.05 was considered statistically significant.

Results
One hundred twelve eyes (54 right and 58 left) of 58 subjects were included in this study. Twenty-nine subjects self-identified as black (mean age 27.5 + /− 4.9 years, 55% male, 45% female) and 29 subjects self-identified as white (mean age 27.8 + /− 4.2 years, 45% male, 55% female). There were no significant differences in the demographics, refractive error, or axial length parameters between the two racial groups (Table 1). There were no significant differences in either the QI (mean QI 8.4 ± 0.98 in black vs 8.6 ± 0.6 in white subjects, p = 0.77) or the SSI (71.9 ± 7.4 in black vs 72.3 ± 8 in white subjects, p = 0.81). Sixteen patients (10 black, 6 white; 32 eyes total) did not have axial length measurements available for analysis. We performed mixed-effects linear regression model analysis of all subjects controlling for two eyes of the same subject and controlling for axial length while excluding the subjects without available axial length measurements. Most of the relationships between macular capillary parameters and race were consistent across these two analyses (Table 2).
Vessel density values in all three capillary plexuses are reported in Table 3. Within the SCP, black subjects showed significantly lower VD values in the fovea (mean VD 18.2 ± 6.01% in black vs 20.7 ± 3.90% in white subjects, p = 0.043) and the parafovea (mean VD 49.9 ± 2.75% in black vs 51.2 ± 2.38% in white subjects, p = 0.014). There was a trend toward lower VD within the SCP of the total area encompassing the fovea and parafovea in    Table 3. Black subjects had significantly larger FAZ area (mean FAZ area 0.332 ± 0.120 mm 2 in black vs 0.221 ± 0.705 mm 2 in white subjects, p = 0.006) and larger FAZ perimeter than white subjects (mean FAZ perimeter 2.22 ± 0.38 mm in black vs 1.91 ± 0.30 mm in white subjects, p = 0.014). There was no significant difference in FAZ AI between black and white subjects (mean FAZ AI 1.11 ± 0.036 in black vs 1.15 ± 0.080 in white subjects, p = 0.076). Vessel density in FD-300 was significantly higher in black as compared to white subjects (FD 53.3 ± 2.84 in black vs 50.3 ± 3.20 in white subjects, p = 0.001).
BFA in the choriocapillaris (Table 3) was significantly lower in black as compared to white subjects in the parafovea (BFA 59.42% ± 1.1 in black vs 61.16% ± 1.3 in white subjects, p < 0.001) and the total 3 mm-diameter circular area centered on the fovea (BFA 60.85% ± 1.1 in black vs BFA 62.78% ± 1.9 in white subjects, p > 0.001). There was a trend toward lower BFA in the fovea in black subjects, however this trend did not reach statistical significance (BFA 60.14% ± 1.7 in black vs 60.81% ± 2 in white subjects, p = 0.11).

Discussion
To date little is known about possible baseline retinal vascular differences across races. To our knowledge, this is the first study evaluating the three retinal capillary plexuses in the normal eyes of black and white young healthy subjects using the latest OCTA software. Our study showed significantly lower VD in the SCP and ICP in the fovea and parafovea in black subjects, as well as a trend toward lower foveal and parafoveal VD in the DCP. We also noted significantly larger FAZ area and perimeter in black subjects. These findings may point toward innate baseline differences in the vascular anatomy of black and white subjects, which may contribute to differing vulnerability to retinal vascular diseases and glaucoma in black patients along with other genetic and socio-economical risk factors previously suggested.
Our earlier study comparing macular capillary parameters in young healthy black and white subjects utilized a previous version of the software segmentation algorithm, which merged the ICP together with the SCP and DCP 20 . The major contribution of this current work is the separate analysis of the three distinct retinal capillary Table 2. Mixed-effects linear regression analysis of the association between race and mean macular capillary parameters. Regression Coefficient and P value accounting for 2 eyes of the same subject. Regression Coefficient* and P* value accounting for 2 eyes of the same subject and adjusting for axial length. VD = vessel density. SCP = superficial capillary plexus. ICP = intermediate capillary plexus. DCP = deep capillary plexus. FAZ = foveal avascular zone. FD-300 = capillary density in a 300 μm wide region around the FAZ. BFA = blood flow area.    www.nature.com/scientificreports/ plexuses. Unlike our earlier study, which showed a significantly lower macular VD within the DCP, and a trend toward lower macular VD within the SCP in black subjects, this study showed a significantly lower macular VD within the SCP and ICP, and a trend toward lower VD within the DCP. We believe that the deviation between our earlier and current results may arise from the differences in VD quantification stemming from improved software for artifact removal and different segmentation slabs, rather than distinctly different anatomical patterns. Upon review of literature, we have noted that in most reports of OCTA parameters in retinal vascular diseases the race of subjects is not specified, which may introduce a confounding bias and affect data interpretation. Our findings of baseline differences in macular capillary parameters among young healthy black and white subjects suggest that these OCTA metrics may vary in populations in ways that are not related to disease. It is important to consider race when analyzing retinal vasculature among different study groups and creating normative databases involving OCTA parameters. Furthermore, as our understanding of the chorioretinal vascular anatomy and physiology evolves with the help of novel imaging modalities, we may choose to select biomarkers that are less susceptible to variation by race in the healthy population.
Black patients with type 2 diabetes mellitus are known to have significantly higher rates of proliferative diabetic retinopathy and diabetic macular edema, and are more likely to experience vision loss and blindness as a result of diabetic retinopathy as compared to non-hispanic white patients [22][23][24][25] . In addition, black patients are more susceptible to the retinal vascular effects of hypertension 22 . Furthermore, among patients with retinal vein occlusions, black patients have more severe visual impairment and require more aggressive treatment as compared to white patients [26][27][28][29] . Better understanding of baseline retinal vascular parameters across races is necessary to understand and address this difference in vulnerability to the damaging effects of sustained hyperglycemia and hypertension.
Several studies analyzing microvascular parameters of the SCP, ICP and DCP in patients with diabetic retinopathy found a significant reduction in vessel density across all three distinct capillary plexuses 30 , with the degree of vessel density reduction corresponding to the severity of diabetic retinopathy and associated vision loss [30][31][32][33][34][35][36] . Recently, the important role of ICP has been distinctly demonstrated with significantly decreased ICP vessel density, perfusion, and blood flow in patients with worsening DR 9,31,37 . Similarly, a significant decrease in macular vessel density within the SCP, ICP and DCP has been demonstrated in patients with BRVO 38 , while the extent of vessel density loss has been shown to correlate with the degree of vision loss following resolution of macular edema 39,40 . VD has been suggested as a potential biomarker of retinal vascular diseases; however there may be a difference in baseline VD within the three distinct capillary plexuses across races as shown in our study. Thus, race needs to be considered as a possible confounder when analyzing macular capillary density Table 3. Macular capillary parameters measured using optical coherence tomography angiography in white and black subjects. Data provided show the results of a mixed effects regression correlating race and macular capillary parameters, adjusting for 2 eyes of the same patient and axial length. Data are provided as mean ± standard deviation. Fovea refers to the central 1-mm diameter circular area. Parafovea refers to the 3-mm diameter area excluding the central 1-mm diameter. P* value calculated controlling for 2 eyes of the same patient, and axial length. VD = vessel density. SCP = superficial capillary plexus. ICP = intermediate capillary plexus. DCP = deep capillary plexus. FAZ = foveal avascular zone. FD-300 = capillary density in a 300 μm wide region around the FAZ. BFA = blood flow area. www.nature.com/scientificreports/ parameters in retinal vascular diseases. Choriocapillaris vessel density in the foveal, parafoveal and perifoveal areas has also been shown to decrease in DR and CRVO in association with ischemic damage 41,42 . Our study showed decreased choriocapillaris BFA in healthy black subjects; this finding needs further investigation in larger age-varied populations, since choriocapillaris BFA may have a role in susceptibility to ischemic damage. Considering that decreased vessel density and increased flow voids in the choriocapillaris have been associated with the development and progression of age-related macular degeneration (AMD), our finding of decreased choriocapillaris BFA in healthy young black subjects is interesting in light of the known lower risk of AMD among black people 43,44 . Differential pigment levels among white and black subjects may lead to different levels of imaging artifact, possibly affecting our ability to uniformly quantify the choriocapillaris. Another explanation could be that increased melanin content may affect choriocapillary vasculogenesis, as some preclinical studies have shown that melanin may have antiangiogenic properties that theoretically could affect choriocapillary density and decrease the risk of AMD 45 . FAZ abnormalities on OCTA imaging have also been associated with retinal vascular diseases. Diabetic retinopathy is associated with enlargement of the FAZ, greater FAZ perimeter, and increased FAZ acircularity index 13 . FAZ enlargement may even represent a pre-symptomatic sign before the clinical onset of diabetic retinopathy 46,47 . Retinal vein occlusions have been associated with enlarged FAZ, increased FAZ acircularity index, a more tortuous pathological FAZ border, and show a negative correlation between visual acuity and FAZ area in patients without macular edema 42,[48][49][50] . Our results indicate a larger FAZ area and perimeter in young healthy black subjects as compared to white subjects. Further validation of this variation is necessary, since baseline differences in FAZ across races may not only affect the validity of this parameter when assessing disease severity, but may also suggest a potential mechanism of greater susceptibility to microvascular changes associated with retinal vascular diseases. The FD-300 value complimentary to other FAZ metrics has been shown to decrease with age in all plexuses and sectors of the macula, and decrease in diabetic retinopathy 51,52 . Our results showed a significantly higher FD-300 density in black as compared to white subjects. This relationship may represent a compensatory mechanism, where the greater capillary density around the foveal avascular zone may compensate for the larger avascular zone and differences in foveal pit morphology in healthy black patients 53 . More studies in larger populations including subjects of different races are necessary to further evaluate FD-300 and its significance.
Recent studies have shown that changes in SCP, which supplies the ganglion cell complex, may play a role in the pathogenesis of open angle glaucoma as well. OCTA imaging studies have shown that patients with open angle glaucoma have a decreased vessel density and a larger area of capillary dropout within the SCP 14,54-56 . Furthermore, increased patient age, a known risk factor for glaucoma development, is shown to be associated with decreased retinal capillary density, however it is unknown whether this trend is different in black as compared to white patients 57,58 . It is unclear whether reduced macular capillary perfusion is contributing to ganglion cell loss in glaucoma, or develops as a consequence of reduced metabolic and vascular demand of already damaged ganglion cells 59 . Our findings of lower baseline vessel density within SCP and ICP in black subjects indicate that race may be a confounder when analyzing the role of superficial retinal vascular complex perfusion in glaucoma development.
The limitations of our study include a small population size, participation of a single clinical site, and engagement of a single image reader. Since our study included only young healthy subjects, these results may not be generalizable to older populations and patients with comorbidities. Our analysis of the chorioretinal capillary parameters is limited by the 3 × 3 mm scan acquisition. Further studies analyzing larger scanned areas can contribute more insights into chorioretinal capillary anatomy and physiology. Additionally, our study utilizes a manual segmentation algorithm, which is limited in its execution by the resolution of the acquired OCTA images and the AngioAnalytics software. The strengths of the study include recruitment of a well-matched patient cohort, and utilization of the latest OCTA software with projection artifact removal allowing analysis of the distinct capillary plexuses.

Conclusions
In conclusion, this study demonstrates decreased vessel density in SCP and ICP, larger FAZ size and perimeter, and decreased choriocapillaris BFA in young healthy black as compared to white subjects. OCTA is a useful imaging modality to study the role of macular microvascular network heterogeneity among racial groups. Further studies including age-varied subjects are needed to validate these macular microvascular differences in larger populations and elucidate any potential clinical associations.