Measuring the foveal avascular zone in diabetes: A study using optical coherence tomography angiography

Abstract Aims/Introduction Diabetes is a global issue that currently affects 425 million people worldwide. One observable microvascular complication of this condition is a change in the foveal avascular zone (FAZ). In this study, we used optical coherence tomography angiography to investigate the effect of diabetes on the FAZ. Materials and Methods A total of 11 participants with diabetes and 11 participants without diabetes took part in this study. Participants in both groups were matched for age (P = 0.217) and sex (P = 0.338), and had no history of ocular disease. Macular optical coherence tomography angiography (OCT‐A) scans of participants’ right and left eyes were taken. Glycosylated hemoglobin (HbA1c) and blood glucose levels were also measured. The FAZ area was manually segmented at the levels of the superficial capillary plexus (FAZSCP) and deep capillary plexus (FAZDCP). Results There was a strong relationship between the FAZ area of participants’ right and left eyes (P ≤ 0.001) in both diabetes and non‐diabetes groups. In the diabetes group, the FAZSCP (P = 0.047) and FAZDCP (P = 0.011) areas was significantly larger than in the non‐diabetes group. Moreover, multiple linear regression analysis predicted a 0.07‐mm2 increase in the FAZSCP and FAZDCP areas of individuals with diabetes for every 1% increase in their HbA1c level. Conclusions Our findings show that there is enlargement of the FAZ in individuals with diabetes compared with individuals without diabetes. In the diabetes group, this enlargement appears to be correlated with HbA1c level. OCT‐A imaging could, therefore, be a useful tool to monitor the FAZ and identify potential early microvasculopathy in diabetes.


INTRODUCTION
Diabetes affects an estimated 425 million people worldwide. In developed countries, diabetic retinopathy (DR) is the leading cause of preventable vision loss in working-age adults aged between 20 and 65 years 1 . Indeed, the presence of DR is approximately 35% in the diabetes population 2 . Consequently, DR has a considerable economic impact 3 .
The retina has high metabolic and oxygen demands; therefore, it is vulnerable to sight-threatening, microvascular complications of diabetes 4 . The central area of the macula is peculiarly susceptible in diabetes because of the foveal avascular zone (FAZ). This region of the human retina, which has the highest cone photoreceptor cell density and provides high-resolution visual acuity, is completely devoid of retinal capillaries 5 . The neurons within the FAZ, like the photoreceptor layer elsewhere in the retina, rely on the blood supply from the choriocapillaris: the superficial capillary plexus (SCP) supplies the retinal nerve fiber layer and ganglion cell layer, whereas the inner retina and outer plexiform layer are supplied by the deep capillary plexus (DCP).
The normal FAZ in a well-developed human fovea is circular when viewed en face 6,7 . The anatomy of the FAZ has been found to vary with age, race, and sex. Previous research has shown that there is a difference between the FAZ in diabetic and healthy eyes 6,8,9 . Furthermore, a recent review found that the diameter of the FAZ in diabetic eyes differs between the capillary plexuses 10 .
Glycosylated hemoglobin (HbA 1c ) is the gold-standard method used to assess long-term glycemic control. Glucose is added to the hemoglobin molecules irreversibly by enzymically catalyzed glycosylation at a rate that is proportional to glucose concentration in the blood. HbA 1c is broken down when these erythrocytes, which have a 3-month lifespan, are destroyed in the liver and spleen. The proportion of hemoglobin that is glycosylated thus serves as a measure of plasma glucose in the preceding 3-month period. The reference range for HbA 1c in an adult without diabetes is 4.0-5.9% 11 , with a value >6.5% being diagnostic of diabetes 12 . Measuring the blood glucose level permits assessment of instantaneous glycemic control in real time. On application of the blood sample to the glucose testing strip, the glucose oxidase enzyme present on the strip interacts with glucose in the sample, taking an electron and forming gluconic acid. The enzyme then passes the electron to water and oxygen, regenerating the enzyme and forming hydrogen peroxide. On glucose testing strips, a mediator replaces oxygen, and this mediator accepts the electron and passes it to an electrode to generate the current that is reported as the glucose concentration 13 . The reference range for blood glucose level in adults is 74-106 mg/ dL 11 , with a value below this range indicating hypoglycemia, and a value above this range indicating hyperglycemia. Participants' glucose levels were recorded with a view to ensuring that it was safe for them to proceed with the study.
Fundus fluorescein angiography (FFA) has been recognized as an important functional imaging technique that provides two-dimensional images of the retina, and permits assessment of blood circulation and vessel integrity. Some diabetic features can be better assessed with FFA than fundus photography. Indeed, FFA can detect primary vascular lesions, such as microaneurysms and intraretinal microvascular abnormalities, and this imaging technique can also identify areas of non-perfusion and neovascularization. However, FFA is limited when studying microvascular histopathological processes: dye-based angiography does not permit separate evaluation of the SCP and DCP, because the capillary plexuses are overlapped when viewed in two dimensions 14 . In addition, the procedure is time consuming and involves the intravenous administration of fluorescein dye with its attendant risks 15 .
The advent of optical coherence tomography angiography (OCT-A) has revolutionized ophthalmic clinical practice. This technology produces high-contrast images of where cells are moving, with sufficiently high resolution to show the locations of individual capillaries in the retina. It differs from FFA, which shows the lumina of retinal blood vessels by making plasma fluorescent, and which shows sites of breakdown of the inner and outer blood-retinal barriers 16 . OCT-A is able to differentiate between the SCP and DCP; therefore, it is possible to show how each plexus is affected in retinal vascular disease. OCT-A is particularly sensitive at detecting early DR 17,18 , insofar as capillary dropout and early retinal neovascularization are well delineated by this method 19 . Indeed, microvascular abnormalities and areas of capillary nonperfusion can also be detected 20 .
The purpose of the present study was to use OCT-A to measure the FAZ area of individuals with diabetes and compare it with that of healthy individuals without diabetes. As HbA 1c level is a marker of progression of diabetes, we also aimed to correlate FAZ with participants' HbA 1c level.

Participants
A total of 22 participants took part in this study, and data were collected from both eyes. All participants were matriculated students at Glasgow Caledonian University or patients who attended the on-campus Vision Centre. Participants were chosen irrespective of their ethnicity and the type of diabetes with which they had been diagnosed. Appointments were arranged to avoid times of day at which participants with diabetes were at a higher risk of becoming hypoglycemic.
In the diabetes group (n = 11), the mean age (-standard deviation) was 37 -17 years, the median age (-interquartile range) was 30 -37 years and the male-to-female ratio was 2:9. In the diabetes group, there were six individuals with type 1 diabetes and five individuals with type 2 diabetes. In the nondiabetes group (n = 11), the mean age was 30 -14 years, the median age was 23 -21 years and the male-to-female ratio was 4:7. These two groups were sufficiently matched for age (U = 42, z = −1.253, P = 0.217) and sex (χ 2 [1, n = 22] = 0.917, P = 0.338) to permit comparison of the diabetes and non-diabetes groups.

Inclusion and exclusion criteria
All participants had a best-corrected visual acuity of 0.3 log-MAR or better in each eye, as measured using a logMAR chart (Thomson Test Chart; Thomson Software Solutions, Hatfield, UK), and the interocular difference in visual acuity was no greater than one line (0.1 logMAR). Participants in the diabetes group had type 1 or 2 diabetes mellitus, as diagnosed by a diabetologist. Furthermore, all participants with diabetes reported no previous diagnosis of DR or diabetic maculopathy, and participants whose fundus photographs showed such diabetic microvasculopathy were excluded. Participants with any concurrent ocular diseasefor example, cataract, age-related macular degeneration or glaucomawere also excluded from the study.

HbA 1c
Participants' HbA 1c levels were measured using the A 1c Now ® + System (PTS Diagnostics, Indianapolis, IN, USA), which used the principle of colorimetry. HbA 1 level was measured immediately prior to OCT-A imaging. A single 5-μL capillary blood sample was obtained using a single-use lancet. Test results were expressed as the percentage of total hemoglobin that was glycosylated in the sample. The method by which the A 1c Now ® + System assesses HbA 1c level has been described previously 21,22 .
Blood glucose Participants' instantaneous blood glucose level was measured using a glucometer (FreeStyle Freedom Lite ® ; Abbott Diabetes Care Inc., Alameda, CA, USA), with a view to ensuring that it was safe for them to proceed with the study. The device was approved for individuals with diabetes to self-test at home. A 0.3-μL capillary blood sample was taken at the same time as obtaining the capillary blood sample that was used for HbA 1c analysis. Results, in mg/dL, were made available to the investigator 5 s after collection of the sample.

OCT-A
The scanning protocol comprised a 4.5 × 4.5-mm macular OCT-A scan (DRI OCT Triton TM ; Topcon, Tokyo, Japan) of participants' right and left eyes. This instrument had a scanning speed of 100,000 A-scans/s and used a wavelength-sweeping laser with a central wavelength of 1,050 nm. An in-built eyetracking system was applied during image acquisition (SMART-Track TM ; Topcon), and proprietary ratio analysis software (OCTARA TM ; Topcon) was used for angiographic processing. En face images were generated of the SCP and DCP (Figure 1), based on automated layer segmentation carried out by the inbuilt digital software (IMAGEnet6 ® ; Topcon Medical Systems, Oakland, NJ, USA). These en face macular images were then exported to an image processing and analysis software package (ImageJ; National Institutes of Health, Bethesda, MD, USA). In essence, each file was converted to an 8-bit image, and a scale was set such that 320 pixels represented 4.5 mm. The Phansalkar method, with a 15-pixel radius, was used for binarization, as previously reported 23 . The FAZ was manually outlined using the polygon tool, and a best-fit oval was generated. FAZ area, in mm 2 , was then automatically calculated by the software at either the level of the SCP (FAZ SCP area) or that of the DCP (FAZ DCP area), depending on the particular image that was being processed (Figure 2).

Statistical analysis
Statistical analyses were carried out using SPSS Statistics 26 (IBM Corp., Armonk, NY, USA). Intraclass correlation analysis was used to assess the strength of any interocular relationship that might have existed between the FAZ area of the right eye and that of the left eye. Two-way mixed analysis of variance (ANOVA) was then used to assess for a difference in the FAZ area, measured in the two distinct capillary plexuses, between the diabetes and non-diabetes groups. A simultaneous method of multiple linear regression was used to examine the effects of HbA 1c level, age, and sex on FAZ area in the diabetes and non-diabetes participants. For all statistical tests, normality of distribution was assessed using Shapiro−Wilk W-tests, and outliers were assessed by inspection of boxplots for values >1.5fold the interquartile range. Furthermore, all parametric assumptions were met, and the alpha-level (α) was set at 0.05.

Interocular relationship
Intraclass correlation analysis showed that there was a strong, statistically significant relationship between the FAZ area of participants' right and left eyes (Table 1). This was the case for measures obtained at the levels of the SCP and DCP in both groups. A two-way random effects model with single measures and absolute agreement was used (intraclass correlation [A,1]) 24 .
In accordance with the statistical guidelines for data obtained from two eyes 25,26 , because there was strong interocular concordance for FAZ area, we used the mean value of the right and left eyes for each participant. In other words, subsequent analysis considered each participant to have two values: (i) the mean FAZ area of right and left eyes, measured at the level of the SCP (FAZ SCP area); and (ii) the mean FAZ area of right and left eyes, measured at the level of the DCP (FAZ DCP area).

Effect of diabetes on FAZ area
A two-way mixed ANOVA was run to examine the effect of diabetes on the FAZ area. The within-subjects' factor was capillary plexus (FAZ SCP area or FAZ DCP area), and the between-subjects' factor was group (diabetes or non-diabetes). Both groups were   Effect of HbA 1c level on FAZ area Multiple linear regression analysis was used to assess the effects of HbA 1c level, sex, and age on FAZ area. In all four of the models described below, a simultaneous method of multiple regression was used.
The first model examined the effect of HbA 1c level, sex and age on the FAZ SCP area of participants with diabetes (Table 2), and the second model examined the effect of HbA 1c level, sex and age on the FAZ SCP area of participants without diabetes (Table 3). In the diabetes group, HbA 1c level significantly predicted FAZ SCP area (β = 0.949, t[7] = 2.828, P = 0.025): there was a 0.07-mm 2 increase in FAZ SCP area for every 1% increase in HbA 1c level ( Figure 4). There was no significant relationship between HbA 1c level and FAZ SCP area in the non-diabetes group (β = 0.192, t[7] = 0.598, P = 0.569). Participants' sex differed between groups: in the diabetes group, male participants had a larger FAZ SCP area compared with female participants (β = 0.152, t[7] = 1.282, P = 0.545), whereas the opposite was found in our non-diabetes group (β = −0.502, t[7] = −1.573, P = 0.169). In the diabetes group, male participants had, on average, a FAZ SCP area that was 0.04 mm 2 larger than that of  The unstandardized regression coefficient is denoted by B; the standardized coefficient is denoted by β; the standardized coefficient divided by its standard error is denoted by t; the coefficient of determination is denoted by R 2 ; the adjusted coefficient of determination is denoted by ΔR 2 . FAZ SCP , foveal avascular zone area manually segmented at the level of the superficial capillary plexus; HbA 1c , glycosylated hemoglobin. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001.  (Table 4), and the fourth model examined the effects of HbA 1c level, sex and age on the FAZ DCP area of participants without diabetes (Table 5). In the diabetes group, the model predicted a 0.07-mm 2 increase in FAZ DCP area for every 1% increase in HbA 1c level (Figure 5); however, this relationship did not reach statistical significance (β = 0.757, t[7] = 1.885, P = 0.101). There was no relationship between FAZ DCP in the non-diabetes group (β = 0.198, t[7] = 0.622, P = 0.545). Likewise, FAZ DCP was not significantly associated with participants' age in either group (diabetes group: β = 0.502, t[7] = 1.342, P = 0.221; nondiabetes group: β = −0.045, t[7] = −0.142, P = 0.891). The effect of sex showed similar findings to those of the FAZ SCP area regression models, but it is of note that these differences did not reach statistical significance in either group: in the diabetes group (β = 0.309, t [21] = 0.928, P = 0.384), male participants had, on average, a FAZ DCP area that was 0.11 mm 2 larger than that of their female counterparts, whereas in the non-diabetes group (β = −0.507, t[23] = −1.601, P = 0.153), male participants had, on average, a FAZ DCP area that was 0.09 mm 2 smaller than that of female participants.

DISCUSSION
The purpose of this study was to measure the FAZ area in individuals with and without diabetes using OCT-A, and relate this measure to participants' HbA 1c level. Our group analysis found that individuals with diabetes had significantly larger FAZ SCP and FAZ DCP areas compared with age-and sexmatched individuals without diabetes. Moreover, our study revealed a positive correlation between HbA 1c level and FAZ area and in individuals with diabetes: for every 1% increase in HbA 1c level, there was a 0.07-mm 2 increase in both FAZ SCP and FAZ DCP areas. This relationship was not present in our cohort without diabetes.
Enlargement of the FAZ is associated with DR, and this finding has been well documented using FFA 6,27,28 . The pathological mechanisms that underlie this enlargement in diabetes are multifactorial: previous evidence has found that capillary closure/dropout 6 , dysfunction of the capillary endothelium 29 , and an increased level of vascular endothelial growth factor 30 might be involved in the pathogenesis. Although many FFA studies have found that diabetic eyes with established DR have an increased FAZ area compared with non-diabetic eyes, only a The unstandardized regression coefficient is denoted by B; the standardized coefficient is denoted by β; the standardized coefficient divided by its standard error (SE) is denoted by t; the co-efficient of determination is denoted by R 2 ; the adjusted coefficient of determination is denoted by ΔR 2 . FAZ SCP , foveal avascular zone area manually segmented at the level of the superficial capillary plexus; HbA 1c , glycosylated hemoglobin. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001. few have included patients with diabetes whose eyes have no or background DR 28,31 . One possibility for this selection bias is that such individuals rarely undergo FFA because of the attendant risks of the procedure. Recent OCT-A studies have also found an enlargement of the FAZ in diabetes 9,32 . The diabetes groups in these studies comprised eyes that had varying degrees of DR. Di et al 32 . found that the FAZ SCP area in diabetic eyes was larger than that in non-diabetic eyes. A similar study used OCT-A to measure the largest diameter of the FAZ in diabetic eyes with DR and in healthy, non-diabetic eyes 9 ; the authors found that the FAZ SCP diameter in diabetic eyes was significantly larger than that in non-diabetic eyes (P = 0.029), with a more pronounced between-group difference being observed in the FAZ DCP diameter (P = 0.001). The OCT-A instrument used in the present study (DRI OCT Triton TM ; Topcon) permitted assessment of the FAZ at the levels of the SCP and DCP, and the sample sizes of our diabetes and nondiabetes groups were equal. Although there is no accepted method to assess the FAZ, we used manual segmentation because of its high repeatability and lower measurement error. The FAZ in all OCT-A images was manually outlined by a single clinician, with a view to eliminating interobserver variability. Furthermore, all images were graded in a randomized order at two points in time, to account for intra-observer variability; the mean value of the two FAZ The unstandardized regression coefficient is denoted by B; the standardized coefficient is denoted by β; the standardized coefficient divided by its standard error is denoted by t; the coefficient of determination is denoted by R 2 ; the adjusted coefficient of determination is denoted by ΔR 2 . FAZ DCP , foveal avascular zone area manually segmented at the level of the deep capillary plexus; HbA 1c , glycosylated hemoglobin. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001. The unstandardized regression coefficient is denoted by B; the standardized coefficient is denoted by β; the standardized coefficient divided by its standard error is denoted by t; the coefficient of determination is denoted by R 2 ; the adjusted coefficient of determination is denoted by ΔR 2 . FAZ DCP , foveal avascular zone area manually segmented at the level of the deep capillary plexus; HbA 1c , glycosylated hemoglobin. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001. areas was used. Other commercially available OCT-A instrumentsfor example, RTVue XR (Avanti; Optovue Inc., Fremont, CA, USA) and Cirrus 5000 HD-OCT (Carl Zeiss Meditec, Inc., Dublin, CA, USA)are semi-automated and use in-built algorithms. Recent evidence has found that these semiautomated instruments tend to overestimate the FAZ area compared with manual segmentation 33 . The size of FAZ is multifactorial and, even in healthy individuals, there is considerable variation in its size. This can pose a challenge when assessing, and making clinical evaluations on, possible pathological enlargement of the FAZ in retinal disease. The FAZ SCP area is smaller than that of the FAZ DCP ; this difference is likely a result of the anatomical differences in the retinal vasculature between the two capillary plexuses. A recent study of FAZ size in a healthy cohort without diabetes found that the mean (-SEM) FAZ SCP area was 0.24 -0.08 mm 2 and the mean FAZ DCP area was 0.38 -0.12 mm 2 [30] . In another study of healthy individuals, the FAZ DCP area was, on average, 0.08 mm 2 larger than the FAZ SCP area 34 . The findings of Tan et al. 34 and Ghassemi et al. 35 mirror those of the present study: in our non-diabetes group, the mean (-SEM) FAZ SCP area was 0.25 -0.03 mm 2 and the mean FAZ DCP area was 0.31 -0.04 mm 2 , with the difference in FAZ between the two capillary plexuses being statistically significant.
We recognize that the present study had some limitations. As the intraclass correlation coefficients between right and left FAZ measures showed a significant interocular relationship, the mean value of FAZ area from both eyes was used 35 . OCT-A studies of healthy eyes have found that interocular measures of FAZ area are similar 25,26 . As this was an explorative study with a relatively small sample size, participants with diabetes were recruited irrespective of the type of diabetes with which they had been diagnosed and irrespective of the duration since their diagnosis. The effect of duration of diabetes on the FAZ remains unclear: an FFA study found that the size of the FAZ increased with stage of DR 36 , whereas a more recent OCT-A study with a similar sample size found that there was no correlation between FAZ size and duration of diabetes 31 . In order to establish the relationship between FAZ area and HbA 1c level, a larger sample size would be required; this would not only increase the statistical power, but it would also allow for the results to be generalized to the diabetes population. The field of view of the OCT-A images captured in the present study was 4.5 × 4.5 mm (approximately 15°), whereas the field of view using FFA is typically ≥30°. It is this difference in field of view that has slowed down the clinical acceptance of OCT-A. Commercially available OCT-A instruments can now produce individual 12 × 12-mm scans; moreover, these scans can be composited using image montage software to produce images with a field of view in the region of 90°3 2 .
The potential for clinical application of the present findings will depend on further longitudinal studies with serial measurement of HbA 1c level and FAZ area in individuals with diabetes.
These additional studies would allow us to establish whether poor glycemic control leads to an increase in the FAZ area, and they would also allow us to understand the effect that disease duration might have on the FAZ area. Additional research would be required to establish whether tight glycemic control could arrest or reverse FAZ enlargement. Adaptive optics methods, which have a higher resolution than OCT-A, offer the potential for better understanding of the pathological mechanisms that are involved in FAZ changes in diabetes 37 .
Although FFA is still considered to be the gold standard in imaging of the retinal vasculature, OCT-A is an evolving technology that can be carried out quickly and non-invasively alongside OCT and digital retinal imaging. As the prevalence of diabetes is projected to rise due to an aging population, early detection of the disease will play a pivotal role, enabling clinicians to better manage and monitor their patients. OCT-A imaging could, therefore, be a useful tool to monitor the FAZ and one that has the potential to identify early microvasculopathy in individuals with diabetes.