Quantitative analysis of the macula with optical coherence tomography angiography in normal Japanese subjects: The Taiwa Study

This study evaluated age-related changes in the superficial and deep retinal capillary plexus (SCP and DCP, respectively) and in the foveal avascular zone (FAZ). SCP and DCP perfusion density (PD) were measured in optical coherence tomography angiography (OCTA) macular scans of 145 eyes of 145 healthy Japanese subjects, and findings were compared with SCP FAZ and clinical data. We found that age was negatively correlated with SCP and DCP PD (r = −0.17, P = 0.04 and r = −0.20, P = 0.02, respectively) and positively correlated with FAZ area (r = 0.18, P = 0.03). SCP and DCP PD were correlated with each other (r = 0.67, P < 0.001). FAZ area was negatively correlated with SCP PD, DCP PD and central macular thickness (CMT) (r = −0.18, P = 0.03; r = −0.25, P < 0.01; and r = −0.39, P < 0.001, respectively). FAZ was larger and CMT was lower (P = 0.01 and P < 0.001, respectively) in women than men. SCP and DCP PD were positively correlated with estimated glomerular filtration rate (r = 0.17, P = 0.03 and r = 0.24, P < 0.01, respectively). Multiple regression analysis confirmed that age independently affected DCP PD and FAZ (P = 0.02 and P < 0.01, respectively) and that CMT independently affected FAZ area (P < 0.001). Thus, normal subjects showed age-related decreases in macular PD and renal function. FAZ and CMT were related, suggesting that age-related changes in macular thickness also affect capillary vasculature.

There was no significant difference in overall SCP PD between the age groups by decade (Fig. 2 left). Overall DCP PD was lower in the 50-59 YO group and the 60-74 YO group than in the 30-39 YO group (P < 0.01 and P < 0.01, respectively; Fig. 2 center). FAZ area in the SCP was larger in the 60-74 YO group than in the 30-39 YO group (P = 0.04; Fig. 2

right).
A comparison of SCP and DCP PD in various areas/quadrants, based on the Early Treatment Diabetic Retinopathy Study (ETDRS), is shown in Table 2. In all subjects, DCP was higher than SCP in the overall, the parafovea, and the superior, nasal, and inferior sectors (P < 0.001, P < 0.001, P < 0.01, P < 0.001 and P < 0.001, respectively), while SCP was higher than DCP in the fovea (P < 0.001). In the 30-39 YO group, DCP was higher than SCP in the overall, the parafovea, and the superior, nasal, and inferior sectors (P < 0.001, P < 0.001, P < 0.001, P = 0.01, P < 0.001 and P < 0.001, respectively), while SCP was higher than DCP in the fovea (P < 0.001). In the 40-49 YO group, DCP was higher than SCP in the parafovea and the nasal and inferior sectors (P = 0.02, P < 0.01 www.nature.com/scientificreports www.nature.com/scientificreports/ and P = 0.02, respectively), while SCP was higher than DCP in the fovea (P = 0.02). In the 60-74 YO group, DCP was higher than SCP in the overall, the parafovea, and the nasal and inferior sectors (P < 0.001, P < 0.001, P < 0.001, and P < 0.01, respectively), while SCP was higher than DCP in the fovea (P < 0.001).
In all included subjects, intra-observer reproducibility of for FAZ area was as follows: intraclass correlation coefficient (ICC) = 0.96 and coefficient of variation (COV) = 5.8%, and inter-observer reproducibility for FAZ area was as follows: ICC = 0.97 and COV = 3.7%.

Discussion
This study establishes a normative database for OCTA measurements of the macular capillary plexus, which promises to allow inter-subject and inter-group comparisons, thereby improving the diagnosis of retinal diseases and glaucoma. We used SS-OCTA to determine PD in the SCP and DCP and FAZ area and calculated the relationship between these results and general findings in healthy Japanese subjects ranging widely in age, all of whom were recruited as part of the Taiwa Study. Our main findings were that SCP and DCP PD were negatively correlated with age, while FAZ area was positively correlated with age. Moreover, SCP and DCP PD showed a www.nature.com/scientificreports www.nature.com/scientificreports/ positive correlation with each other, while FAZ area showed negative correlations with SCP PD, DCP PD and CMT. FAZ was larger and CMT was lower in the female subjects than the male subjects. SCP and DCP PD were both positively correlated with eGFR. Multiple regression analysis confirmed that age was an independent factor affecting DCP PD and FAZ and that CMT was also an independent factor affecting FAZ area.

Comparison with previous normative data for SCP and DCP PD.
There have been only a few studies discussing OCTA-measured VD, all of which used circular regions of interest (ROI) in the normal macula 17,18,[21][22][23] . In one such previous study, PD in the parafoveal SCP and DCP was reported to be 46.0% and 51.5%, respectively 22 , similar to our findings of 48.2% and 50.0%. However, the same study reported that PD in the foveal SCP and DCP was 31.9% and 27.5%, respectively 22 , which is much higher than our findings. This might be . SCP and DCP PD and FAZ in the four groups classified by age. There was no significant difference in SCP PD between the four groups (left). DCP PD was lower in the 50-59 YO group and the 60-74 YO group than in the 30-39 YO group (P < 0.01 and P < 0.01, respectively; center). FAZ area in the SCP was larger in the 60-74 YO group than in the 30-39 YO group (P = 0.04; right).

Overall (%) Fovea (%) Parafovea (%) Temporal (%) Superior (%) Nasal (%) Inferior (%)
All ( www.nature.com/scientificreports www.nature.com/scientificreports/ explained by the fact that the previous study defined the foveal region as a central, 1.2-mm circle comprising 120 pixels, in contrast with the current study, which used a 1.0-mm circle. Circular ROIs for VD are the most suitable to compare VD in different sectors of the macula, considering the anatomy of the macular vascularity and the fact that the distance from the foveal center is fixed. The current study, which also used a circular ROI, adds support to existing findings that DCP PD is higher than SCP PD in the parafovea, but, inversely, is lower in the fovea 21,22 . Higher DCP PD in the parafovea may be explained by the composition of the SCP and DCP. Specifically, the SCP includes transverse capillaries and the DCP includes the homogenous capillary vortex 23,24 . Furthermore, interestingly, our study found no significant differences in SCP and DCP PD in the temporal sector, regardless of age. This might be because DCP PD in this sector is relatively low, possibly due to the specific anatomical vascularity of the temporal raphe 25 , a horizontal boundary separating the superior and inferior retinal nerve fiber bundles in the temporal retina, i.e., the watershed zones. Our study also showed a positive correlation between SCP and DCP PD in the overall area. Considered together, these findings show that in normal eyes, there is a close association between SCP and DCP PD, and that SCP PD is higher in the foveal area, while DCP PD is higher in the parafoveal area.    www.nature.com/scientificreports www.nature.com/scientificreports/ Aging. Generally, the health of the human vascular system is closely associated with age. The current study demonstrated that SCP and DCP PD were negatively correlated with age, while SCP FAZ area was positively correlated with age. This is consistent with previous studies showing similar relationships between changes in the macular capillary plexus and age 17,18,22,23 . These age-related changes have been reported to be caused by occlusion and atrophy of the retinal capillaries 26 . However, in contrast to DCP PD, SCP PD was similar in the four groups in the current study, and furthermore, a multivariate analysis confirmed that only DCP was closely associated with age (with a per-year decrease of 0.23%). Although a previous study that investigated human donor eyes with confocal microscopy found no age-dependent changes in retinal capillary density 27 , previous studies using methods more similar to ours found annual decreases in the macular capillary plexus and annual increases in FAZ 17,18 . Thus, we speculate that DCP PD decreases and SCP FAZ increases with age, probably because of reductions in oxygen and nutrient demand, in the normal population.
We observed only weak tendencies in the data from the middle-aged subjects (i.e., the 40-49 YO and 50-59 YO groups), probably due to the small numbers of subjects in these groups. These small group sizes were possibly due to lifestyle factors, which do not allow middle-aged individuals to attend annual medical checks.

Sex.
Male sex has previously been reported to be associated only with increased SCP PD 28 . In our study, though there were sex differences in some background characteristics, including blood pressure and hepatic and biliary enzymes (which also differed somewhat in the younger and older groups), our results are compatible with previous reports that there were no sex differences in PD 21,22,29,30 . Furthermore, the central subfield and foveal and parafoveal thickness have been reported to be lower in women than men 20,31 . Our finding that CMT was lower    Table 4. Multiple regression analysis of factors independently contributing to OCTA parameters. PD = perfusion density, IOP = intraocular pressure, SBP = systolic blood pressure, HbA1c = hemoglobin A1c, CMT = central macular thickness, SCP = superficial retinal capillary plexus, DCP = deep retinal capillary plexus, FAZ = foveal avascular zone. β = standard partial regression coefficient.
www.nature.com/scientificreports www.nature.com/scientificreports/ in the female than the male subjects confirms that this tendency is present regardless of age. Women have also been reported to have a larger superficial and deep FAZ 32 . Similarly, our results showed that, in all subjects and in the younger subjects, FAZ was larger in the female than the male subjects. There were no sex differences in FAZ among the older subjects, possibly because axial length was longer in the male than in the female subjects. Considered together with past findings, we speculate that there are no sex differences in PD, although there are in CMT and FAZ.

FAZ.
We found that FAZ area increased with age 18,33 . Enlarged FAZ (a novel OCTA parameter) has also previously been reported to be associated with various ocular diseases, including diabetic retinopathy and retinal vein occlusion 34,35 . Since the border of the FAZ is more clearly visible at the level of the SCP than the DCP 36,37 , many investigators have focused on SCP FAZ [37][38][39][40][41] . Reproducibility has also been reported to be relatively low in measurements of DCP FAZ. Intra-and inter-observer agreement are both better in SCP FAZ than in DCP FAZ 42 . Furthermore, studies of the macular capillary plexus that use manufacturer-recommended default settings might be biased 43 , and we speculate that measurements of DCP FAZ may be especially unreliable because of stronger errors in segmentation of the retinal layers when the default settings are used.
We found that FAZ in the SCP measured 0.37 mm 2 , which is comparable to previous results for FAZ (0.30~0.35 mm 2 ), measured both with OCTA and with other approaches such as fluorescein angiography and scanning laser ophthalmoscopy [37][38][39][40][41] . We also found that FAZ area was negatively correlated with SCP and DCP PD, which is also consistent with previous findings 21 . FAZ has previously been shown to be larger in eyes with a deeper and broader foveal pit 44 , while a small FAZ has been shown to be a historic mark of prematurity 45 . In pediatrics, a larger FAZ was significantly associated with older age and with reduced foveal macular thickness 28 . Furthermore, our multivariate analysis showed that FAZ was negatively correlated with CMT, again agreeing with previous results 21,36,46 . Thus, current and previous findings indicate that SCP FAZ area increases significantly as CMT decreases. A positive correlation between inner retinal thickness and parafoveal PD has also been reported in healthy subjects 47 . Therefore, considering all results together, we consider that parafoveal vascularity is most likely closely associated with macular thickness.
Associations between OCTA and general findings. The current study is the first to evaluate the association between SS-OCTA parameters and general findings, including renal function. Though renal function can already be evaluated based on eGFR, we found, interestingly, that overall SCP PD, overall DCP PD, and parafoveal   Table 5. Characteristics of eyes divided by age and sex. SBP = systolic blood pressure, DBP = diastolic blood pressure, TG = triglyceride, AST = aspartate aminotransferase, ALT = alanine aminotransferase, γ-GTP = γ-glutamyl transpeptidase, eGFR = glomerular filtration rate, HbA1c = hemoglobin A1c, IOP = intraocular pressure, CMT = central macular thickness, SCP = superficial retinal capillary plexus, PD = perfusion density, DCP = deep retinal capillary plexus, FAZ = foveal avascular zone. a Chi-square test.
www.nature.com/scientificreports www.nature.com/scientificreports/ DCP PD were positively correlated with eGFR. We included 6 different blood tests, but eGFR was the only one associated with both types of PD. This may be explained by past reports showing a close relationship between renal function and ocular circulation [48][49][50] . In older subjects (more than 60 YO), eGFR has been reported to be independently related to central retinal arteriolar equivalent 49 . The progression of chronic kidney disease has also been reported to be significantly and independently associated with decreased retinal blood flow, measured by a laser Doppler velocimetry system, in early-phase diabetic retinopathy 50 . Furthermore, in a study using static and continuous retinal vessel responses to three cycles of flickering light, eGFR was found to be linked to arterial reaction time, arterial maximum dilatation and the dilatation amplitude responses in diabetes and cardiovascular disease 48 . Thus, since there might be a close relationship between renal function and ocular circulation, renal function must be considered when interpreting retinal circulation.
Limitations and advantages. The current study was somewhat limited by its cross-sectional design.
Additionally, we did not use a third, intermediate retinal capillary plexus, but only SCP and DCP. The number of middle-aged subjects was also relatively small, because of the restricted study design. Another possible limitation is that we used age, but not eGFR, in our multiple regression analyses. We made this choice because both the present and previous reports have shown that age and eGFR are closely correlated (current study: r = −0.56, P < 0.001) [51][52][53] .
This study also had many advantages, including a sample size that was large overall, included only one eye from each subject, included only a single ethnicity, and had a wide age range. Reproducibility in the measurements of FAZ was also very high (intra-observer and inter-observer reproducibility: both ICC > 0.95 and both COV < 6%). The study also used strict inclusion criteria to define healthy subjects, confirmed by blood tests and the decisions of medical doctors. Furthermore, to reduce axial length-related data bias, eyes with an axial length greater than 26.5 mm were excluded from this study. Mean image quality was more than 60 in all groups, suggesting that our analysis had good quality. There were no significant differences between most of the groups (with the exception of the youngest and oldest groups) likely due to senile cataract.

Conclusion
In conclusion, we found that macular capillary plexus parameters, measured with SS-OCTA, were closely associated with age and general findings, and found that SS-OCTA was an excellent way of obtaining data on retinal vascularity in separate layers of the macular capillary plexus. Our key findings were that SCP and DCP PD were negatively correlated with age, while FAZ area was positively correlated with age. Additionally, SCP and DCP PD showed a positive correlation with each other. FAZ area showed a negative correlation with SCP PD, DCP PD and CMT, and FAZ was larger, while CMT was lower, in the female subjects than the male subjects. Finally, SCP and DCP PD were both positively correlated with eGFR. Thus, our findings indicate that macular capillary vascularity decreases with age, in association with declining renal function, in normal subjects. Moreover, the relationship between FAZ area and CMT suggests that changes in age-related macular thickness occur in close association with changes in capillary vasculature. We believe that the current results, obtained from normal subjects, should be helpful in improving clinical diagnoses and should allow the comparison of macular capillary plexus parameters obtained with SS-OCTA. Methods setting and design. This was an institutional, cross-sectional case series. patients. Recruited from the Taiwa Study; 924 eyes of 462 subjects. All subjects attended an annual medical check in the town of Taiwa, northern Japan, on the 26 th and 29 th of May or the 3 rd and 4 th of June in 2017. Subjects were excluded if they had a medical history of glaucoma, obviously abnormal fundus findings, decimal best-corrected visual acuity < 0.7, visual field abnormalities, axial length more than 26.5 mm, diabetes (history of medical treatment or HbA1c ≧ 7.0), or hypertension (SBP ≧ 160). Finally, this study included 145 eyes of 145 healthy Japanese subjects who underwent 3 × 3 mm OCTA scans of the macula as part of the Taiwa Study (Fig. 1).
The Measurement of macular capillary plexus and thickness. Subjects underwent 3 × 3 mm OCTA scans (SS-OCT Angio, Topcon Corporation, Tokyo, Japan) of the macula. We defined the PD as the total area of perfused vasculature per unit area in a region of measurement, in conformity with a previous paper 54 . PD was www.nature.com/scientificreports www.nature.com/scientificreports/ calculated in a fovea-centered 2.5-mm diameter circle (whole retina). PD was measured in an inner, 1.0-mm diameter ring around the foveal center, (i.e., the fovea), a larger, 2.5-mm diameter ring (i.e., the parafovea), and in the four quadrants (i.e., temporal, superior, nasal and inferior) PD calculation was based on the ETDRS. We measured the PD in the SCP and DCP, and FAZ area in the SCP, and determined the relationship between these parameters and other clinical findings, including age and the results of blood testing. We use customized third-party software (IMAGEnet6 version 1.23) to remove the large vessels. The resulting OCTA images, used to calculate the vessel density, were not skeletonized, but only binarized. We also determined the correlation between CMT and these parameters. To determine CMT, 7 × 7 mm OCT cube scans centered on the fovea were made. CMT was automatically calculated by averaging the retinal thickness of the macula within 1 mm of the fovea. All included SS-OCT data had a minimum image quality > 35; images were excluded if automatic layer segmentation failed or we judged an image as having significant artifacts.
Automatic analysis software (Topcon Corporation, Tokyo, Japan) was used to set a dividing line 15.6 μm below the boundary between the inner plexiform layer and inner nuclear layer in the macula (i.e., below the SCP and above the DCP). This novel system automatically identified the foveal center and measured SCP and DCP within a 2.5-mm circle. FAZ area in the SCP was calculated manually with ImageJ software. We then determined the statistical associations between these measurements.
OCTA data can be corrected for axial length, but the effects of this correction are still open to debate. Furthermore, we consider that correction might not be the best choice for evaluating the capillaries of the macula. Therefore, in this study, we decided to use uncorrected OCTA data Reproducibilty. The repeatability and reproducibility of the VD results obtained with this software were not investigated. However, the ICC and the COV were calculated to determine the intra-and inter-observer reproducibility of the manually determined data for FAZ area. The intra-observer reproducibility was determined by analyzing results from one observer, who performed the measurements twice. The inter-observer reproducibility was determined by analyzing results from two observers. statistical analyses. The data are presented as mean ± standard error of the mean. The subjects were divided into age groups by decade (i.e., 30-39 YO, 40-49 YO, 50-59 YO and 60-69 YO); the subjects were also divided into younger and older groups, comprising those 30-59 YO and 60-69 YO, respectively. The Chi-square test was used to evaluate differences in sex between the age groups. A one-way analysis of variance was used to evaluate differences in SBP, DBP, TG, AST, ALT, γ-GTP, eGFR, HbA1c, axial length, IOP, CMT, SCP PD, DCP PD and FAZ area in the age groups. Spearman's rank correlation test was used to estimate the relationships of SCP PD, DCP PD and FAZ area to SBP, DBP, TG, AST, ALT, γ-GTP, eGFR, HbA1c, axial length, IOP, CMT, SCP PD, DCP PD and FAZ. The mean image quality of SS-OCTA by age group was compared with the Tukey-Kramer test. Separate multiple linear regression analyses were performed to analyze each independent variable potentially affecting SCP PD, DCP PD and FAZ. All statistical analyses were performed with JMP software (Pro version 10.0.2, SAS Institute Japan Inc., Tokyo, Japan). Differences were considered significant at P < 0.05.