Longitudinal Structure–Function Relationship Between Macular Vessel Density and Thickness and Central Visual Field in Early Glaucoma

Abstract

Purpose.

Investigate the relationship of longitudinal changes in macular vessel density (VD) (from optical coherence tomography-angiography-OCTA) and ganglion cell complex (GCC) (from OCT) with central visual field (VF) in early glaucoma eyes.

Design.

Observational cohort.

Participants.

95 eyes, 37 preperimetric and 58 early glaucoma (24–2 VF mean deviation (MD)≥–6 dB), with average follow–up of 3.8 years and 5.3 visits were included.

Methods.

Whole image (wiVD and wiGCC) and parafoveal scans, as well as localized regions of interest (LROI), hemiretinas of whole image and superior, inferior, temporal and nasal sectors of parafoveal maps, were matched with central VF locations. Age-adjusted rates of change of VD, GCC, mean sensitivity (MS) of VF locations and 10–2 VF MD were calculated with linear-mixed–effect models. Normalized rates of change were calculated for comparing between change rates in wiVD and wiGCC.

Main Outcome Measure.

Structure-function (SF) correlations between VD/GCC and central VF measurement change rates and comparison between the correlations of SF relationships after bootstrapping the difference of the correlations.

Results.

VD loss and GCC thinning demonstrated significant correlations with central VF damage, globally and with most LROIs. The SF correlation (r, 95% confidence interval (CI)) was 0.42 (0.24, 0.58) between wiVD and 10–2 VF MD change rates and was 0.27 (0.08, 0.45) between wiGCC and 10–2 VF MD changes rates, all P < 0.05. In contrast to GCC thinning, VD loss in the parafoveal sectors demonstrated significant correlations with central VF damage in inferior and temporal sectors. Differences between the SF relationship with central VF damage was not significant between VD loss and GCC thinning. The mean (95% CI) of normalized change rates of wiVD (−7.40 ((−7.71, −7.09)%/year) was faster than wiGCC (−1.95 ((−2.21, −1.70)%/year), P<0.05.

Conclusion.

Rates of VD loss and GCC thinning are associated with central VF loss over time. Assessment of both macular VD and GCC thickness should be considered for evaluation of glaucoma progression.

Precis

Macular vessel density and ganglion cell thickness demonstrated a significant longitudinal structure-function relationship with the central visual field damage over time, in early glaucoma. Evaluation of both should be considered for monitoring of glaucoma progression.

Introduction

Glaucomatous damage is characterized by loss of retinal ganglion cells (RGC) which results in progressive functional loss with loss of visual field (VF).^1,2 Spectral-domain optical coherence tomography (SD–OCT) of the macular is used to assess the central RGC–axonal complex.^3,4 In addition to the damaging effects of intraocular pressure, there is considerable evidence for alterations of vascular blood flow in the pathogenesis of glaucoma.^5–7 Impaired blood flow results in loss of RGCs and thinning of the ganglion cell layer. Optical coherence tomography angiography (OCTA) provides a noninvasive and repeatable evaluation of vasculature in the optic nerve and macula.^8,9

Evaluation of change of vessel density (VD) with OCTA is a promising approach for the detection of glaucoma progression.^10–13 A strong correlation between VD dropout and the level of glaucoma severity has been shown in several studies.^14–17 However, these studies have evaluated cross-sectional relationships between VD dropout and VF global indices. One study showed a moderate correlation between macular VD dropout and localized central VF locations¹⁸ However, prior studies have not examined the longitudinal structure-function (SF) relationship between macular VD loss and visual field parameters.

A number of studies have documented central VF loss in early glaucoma.^19–25 It has been reported that both macular OCT and central VF are beneficial for detection of glaucomatous defects in early glaucoma.²⁶ In a recent study by Leung et al., it has been shown that papillofoveal and papillomacular bundles had defects in early glaucoma and the damage was associated with damage in the corresponding VF locations.²⁷ In their study, only 25.0% of eyes with early glaucoma had arcuate bundle defects without involvement of papillomacular and papillofoveal bundles. Garg et al. reported higher rates of glaucoma progression in eyes that demonstrated central VF damage.¹⁹ In another study, a stronger association between central VF and quality of life was reported in comparison to peripheral VF.²² The purpose of the current study was to investigate the longitudinal SF relationship between macular VD loss over time and central VF measurements in the early stages of glaucoma.

Methods

Participants

Ninety–five eyes from 70 patients, 37 preperimetric and 58 early primary open angle glaucoma (POAG) patients from the Diagnostic Innovations in Glaucoma Study (DIGS) were enrolled in this study.^28,29 Patients underwent serial OCTA and macular SD–OCT (Angiovue; Optovue Inc. Fremont, CA, USA) images, and central VF exams^8,9,30–35. The protocol of the study was approved by institutional review board of University of California, San Diego. Written informed consent was obtained from all participants and the methodology of the study was adherent to Declaration of Helsinki and the Health Insurance Portability and Accountability Act.

Eligible patients were required to have at least two years of follow up with a minimum of 4 OCTA and macular SD–OCT images and central VF exams. Inclusion criteria also included older than 18 years of age, open angles defined by gonioscopy, and a best–corrected visual acuity of 20/40 or better at study entry. Exclusion criteria was history of intraocular surgery (except for uncomplicated glaucoma and cataract surgery), secondary causes of elevated intraocular pressure (IOP), other intraocular eye disease, and axial length of 27 mm or more. Participants with the diagnosis of systemic diseases such as Parkinson’s disease, Alzheimer’s disease, dementia, or a history of stroke were also excluded.

The criteria for POAG patients, based on 24–2 VF damage, were as follow: glaucoma hemifield test outside normal limit and pattern standard deviation outside less than 5% of normal on at least 2 consecutive exams.³⁶ The reliability criteria for VF exams were false positive (FP) $\le$ 15% and fixation loss (FL) and false negative (FN) $\le$ 33%. Early glaucoma was defined as baseline 24–2 mean deviation (MD) $\ge$ −6. Preperimetric glaucoma included eyes with elevated IOP (≥22mmHg) or glaucomatous-appearing optic discs without the presence of repeatable glaucomatous VF damage. Glaucomatous-appearing optic disc was defined as notching, neuroretinal narrowing, excavation, localized or diffuse retinal nerve fiber layer defect^10,28

OCTA and SD–OCT imaging

Optical coherence tomography angiography and macular SD–OCT were acquired with AngioVue (Optovue, Inc, Fremont, CA). This imaging system (software version 5.6.3.0) is composed of both angiography and SD–OCT platform and therefore in the same examination, it provides both vascular measurement parameters and thickness of macula between internal limiting membrane (ILM) and inner plexiform layer (IPL), known as GCC.

Macula 3×3–mm² scans, centered on fovea and composed of 304 B–scans × 304 A–scans per B–scan, were obtained with AngioVue imaging system. To calculate VD, split–spectrum amplitude– decorrelation angiography (SSADA) method is applied on the scan slab. The AngioVue software is equipped with automatic segmentation and superficial retinal capillary, located between ILM and IPL. The SSADA capture the dynamic motion of the red blood cells and provide a high–resolution 3–dimensional visualization of perfused retinal vasculature. In the next step, VD is calculated as the percentage of area occupied by blood vessels in any specific location. For this study, the VD 3×3–mm² whole image (wiVD) and parafoveal VD (pfVD) maps were analyzed. The localized regions of interest (LROI) were the superior and inferior hemiretina of the wiVD, map and the standard instrument defined superior, inferior, temporal and nasal sectors of pfVD map.

The same macular cube provides thickness measurements of GCC. The areas of interest were the 3×3–mm² whole image GCC (wiGCC) and parafovea GCC (pfGCC) maps and the LROI described above, the superior and inferior hemiretina of the wiGCC map and superior, inferior, temporal and nasal sectors of pfGCC map.

The quality of OCTA and SD–OCT images were reviewed according to the protocol of the Imaging Data Evaluation and Analysis Reading Center on all the scans which were processed by the AngioVue software.³⁷ Trained reviewers evaluated the scans and images and excluded the poor–quality images with the following definitions: 1) scan quality less than 4; 2) poor clarity images; 3) residual motion artifacts visible as an irregular vessel; 4) weak signals in local areas due to media opacity; 5) uncorrectable segmentation errors.

Visual field testing

Central VF was tested with a 10–2 pattern of Humphrey Field Analyzer (Carl Zeiss Meditec^®, Dublin, CA). As a matching protocol, VF visits that were within six months of OCTA and SD–OCT visits were selected. The reliability criteria were FL and FN $\le$ 33% and FP $\le$ 15%. This software measures VF sensitivity at 68 locations in central 10° which are 2° apart vertically and horizontally. The quality of the visual field tests was reviewed by the Visual Field Assessment Center (VisFACT). The VF data were exported as an XML file and 10–2 MD and threshold sensitivity for each location were extracted.

Mapping OCTA VD and SD–OCT GCC maps and 10–2 VF measurements

The 3×3–mm² scans approximately cover the central 9° of the macula. The correspondence of structural and functional parameters was based on the central VF locations that occupy a specific area of the scan, after adjusting for RGC displacement, as shown in figure 1.^18,38,39 The VF locations were flipped vertically so that they matched the corresponding anatomic areas in the presented images. A total of 8 central locations corresponded to the wiVD and wiGCC maps (4 VF locations per each hemiretina), while for the parafovea map, 4 central locations were matched, with one location per each sector (Figure 1). To calculate the mean sensitivity (MS) of the cluster of VF locations, the threshold sensitivity of each location was converted to the lambert scale and consequently was averaged into the MS of those locations. The averaged MS value was converted back to a logarithmic scale for further analysis.

Figure 1.

Mapping 3×3–mm² whole image vessel density (VD) and ganglion cell complex (GCC) (A) and parafoveal VD and GCC map with central visual field locations, after adjustment for the displacement of retinal ganglion cells.

Statistical analysis

Demographic characteristics were reported as mean (95% confidence interval (CI)) for continuous variables and count (percentage) for categorical variables. Rates of change of VD, GCC and VF MD and MS/threshold sensitivity at each indicator were investigated with linear mixed effect, with random eye–patients intercept and independent random slopes within eyes. This model provides the ability to obtain the best linear unbiased prediction (BLUP).^40,41 The BLUP accounts for individual measurements and estimates of the linear mixed effect which is applied to the entire sample. This method represents inter-subject heterogeneity by using random slopes and intercepts to introduce subject deviation from the average dataset.⁴²

The SF correlation between OCTA and OCT parameters and VF measurement was evaluated with the parametric method. Structure-function correlations will be reported as Pearson’s “r” and the corresponding 95% CI and P-value.

To investigate the difference between the SF relationship between VD loss and central VF damage and the SF relationship between GCC thinning and central VF damage, the difference of correlations was performed with bootstrapping 10000 times.⁴³ This approach provides a normal distribution of the difference between the SF correlations and the average difference along with 95% CI was reported. The difference in the SF correlations was provided for both global indices and the LROI. If zero was included in the 95% CI, the difference between the SF correlations was statistically nonsignificant.

Comparison between rates of change of VD and GCC for this group of POAG eyes at the whole and parafovea images were evaluated using a linear mixed-effect model. For this purpose, since the measures differ in the range of values and units, dynamic range–based normalized coefficients from VD and GCC thickness were estimated.⁴⁴ In this method, the means of the lowest and highest 3% (quantiles) of eyes were calculated to define the dynamic range. The percent of dynamic range change, i.e., normalized variable, was then calculated using the following formula: [(visit value –floor value) / dynamic range] × 100. Age-adjusted normalized slopes were measured through the linear mixed effect model described above. Simple linear regression between the normalized slopes of VD and GCC was performed and R² value and P value were reported.

Statistical analyses were performed using statistical software R version 4.0.4 (R Foundation for Statistical Computing, Vienna, Austria).⁴⁵ P values less than 0.05 were considered statistically significant.

Results

A total of 95 eyes from 70 patients with early-stage glaucoma were included in this study. Table 1 represents the demographic characteristics of the study eyes. The mean (95% confidence interval (CI)) of follow-up time and the number of visits were 3.8 (3.7, 4.0) years and 5.3 (5.0, 5.6) visits, respectively. The average (95% CI) baseline 10° VF MD and 24–2 MD were −1.1(−1.6, −0.7) dB and −1.4(−1.7, −1.0) dB, respectively. Eighty-one eyes (85%) received at least one IOP lowering medication during the follow-up period.

Table1.

Baseline demographic characteristics of study eyes.

Characteristics	N=70, Eyes=95
Age (years)	68.3 (68.1, 68.6)
Gender (%)
Female / Male	57 (60%) /38 (40%)
Race (%)
African American/ Non-African American	28 (30%) / 67 (70%)
Baseline IOP (mmHg)	15.7 (15.6, 15.8)
Baseline CCT (μm)	539.0 (538.1, 539.9)
Diagnosis (%)
POAG / Glaucoma suspect	58(61%)/37(39%)
Follow up time (years)	3.8 (3.7, 4.0)
Number of exams	5.3 (5.0, 5.6)
Baseline 24–2 MD (dB)	−1.4 (−1.7, −1.0)
Baseline 10–2 MD (dB)	−1.1 (−1.6, −0.7)
Baseline wiVD (%)	45.7 (44.8, 46.5)
Baseline wiGCC (μm)	92.0 (89.8, 94.2)
10–2 MD rates of change (dB/y)	−0.21 (−0.2, −0.1)
wiVD rates of change (%/y)	−1.0 (−1.1, −1.0)
wiGCC rates of change (μm/y)	0.8 (−0.8, −0.7)

IOP = intraocular pressure; CCT = central corneal thickness; POAG = primary open angle glaucoma; MD = mean deviation; wiVD = whole image vessel density; wiGCC = whole image ganglion cell complex. Values are shown in mean (95% confidence interval), unless otherwise indicated.

For the global parameters, the mean (95% CI) age-adjusted rates of change of wiVD, wiGCC, pfVD and pfGCC were −1.02 %/year (−1.08, −0.97), −0.93 μm/year (−1.14, −0.72), −0.94 %/year (−1.00, −0.88) and −0.96 μm/year (−1.20, −0.72), respectively (Table 2). The LROIs of the whole image and parafoveal VD map had a range of average change rates between −1.37 %/year (parafoveal temporal sector) and −0.36 %/year (parafoveal nasal sector) and for the GCC parameters, the range of average change rates was between −1.04 μm/year (parafoveal nasal sector) and −0.84 μm/year (parafoveal temporal sector).

Table 2.

Age adjusted rates of change of vessel density (VD), ganglion cell complex (GCC) and central visual field mean sensitivity estimated through linear mixed effect model. The mean and 95% confidence interval (CI) are represented for global and sectoral indicators of VD and GCC maps. (N=70 patients, 95 eyes)

	Vessel Density rates (%/year)	Central Visual Field (dB/year)	Ganglion Cell Complex (μm/year)
	Mean (95% CI)	Mean (95% CI)	Mean (95% CI)
Whole Image map	−1.02 (−1.08, −0.97)	−0.14 (−0.18, −0.09)	−0.93 (−1.14, −0.72)
Superior Hemiretina	−0.76 (−0.82, −0.70)	−0.04 (−0.09, 0.02)	−0.96 (−1.19, −0.73)
Inferior Hemiretina	−1.36 (−1.42, −1.31)	−0.32 (−0.39, −0.26)	−0.92 (−1.11, −0.72)
Parafovea map	−0.94 (−1.00, −0.88)	−0.23 (−0.27, −0.18)	−0.96(−1.20, −0.72)
Superior Sector	−0.90 (−0.97, −0.83)	−0.03 (−0.08, 0.02)	−1.00 (−1.26, −0.73)
Inferior Sector	−1.14 (−1.22, −1.06)	−0.77 (−0.84, −0.69)	−0.96 (−1.21, −0.72)
Temporal Sector	−1.37 (−1.44, −1.30)	−0.05 (−0.13, 0.03)	−0.84 (−1.07, −0.62)
Nasal Sector	−0.36 (−0.42, −0.31)	−0.02 (−0.12, 0.07)	−1.04 (−1.27, −0.80)

CI= 95% confidence interval.

Table 3 summarizes the correlation for structural and functional rates of change at global whole image vessel density, parafoveal vessel density and GCC maps and the LROIs. For the whole image vessel density map, the correlation coefficient (r, 95% CI) between wiVD change rates and 10–2 VF MD change rates was 0.42 (0.24, 0.58), P < 0.001, (Figure 2 A). For the LROIs, the topographic correlation (r, (95% CI)) between wiVD change rates and MS change rate of the central 8 VF locations was 0.26 (0.06, 0.44), P = 0.013, while the SF correlation (r (95% CI)) for superior hemiretina was 0.32 (0.13, 0.49), P = 0.001 and for inferior hemiretina was 0.31 (0.12, 0.48), P = 0.001. The correlation (r (95% CI)) of the pfVD map change rates and MS change rate of the 4 central locations was 0.21 (0.01, 0.39), P = 0.036. The SF relationship of the LROI of the parafoveal map was significantly higher in inferior sector ((r (95% CI)) 0.24 (0.04, 0.42), P = 0.018) compared to superior sector (0.14 (−0.05, 0.34), P = 0.149). There were similar SF correlations (r (95% CI)) for temporal (0.23 (0.03, 0.41), P = 0.022) and nasal (0.23 (0.03, 0.41), P = 0.024) sectors.

Table 3.

Structure-function (SF) correlations (r) between central visual field (VF) defect and age-adjusted vessel density loss and ganglion cell complex thinning. Differences (95% CI) between the two SF relationships are calculated with Bootstrapping methods. The differences between the corresponding SF relationships were considered not significant when the 95% CI contains zero. (N=70 patients, 95 eyes)

	Vessel Density Loss		GCC Thinning		Difference in r (95%CI)
	r Coefficient (95% CI)	P value	r Coefficient (95% CI)	P value
Whole image and 10–2 MD	0.42 (0.24, 0.58)	< 0.001	0.27 (0.08, 0.45)	0.005	0.14 (−0.29, 0.59)
Whole image and VF MS	0.29 (0.09, 0.46)	0.004	0.29 (0.10, 0.47)	0.004	0.00 (−0.30, 0.29)
Superior Hemiretina	0.32 (0.13, 0.49)	0.001	0.43 (0.26, 0.58)	<0.001	−0.11 (−0.49, 0.17)
Inferior Hemiretina	0.31 (0.12, 0.48)	0.001	0.08 (−0.11, 0.28)	0.410	0.23 (−0.17, 0.71)
Parafovea and VF MS	0.21 (0.01, 0.39)	0.036	0.22 (0.02, 0.41)	0.026	−0.01 (−0.34, 0.28)
Superior Sector	0.14 (−0.05, 0.34)	0.149	0.25 (0.05, 0.43)	0.011	0.10 (−0.40, 0.15)
Inferior Sector	0.24 (0.04, 0.42)	0.018	0.10 (−0.1, 0.3)	0.323	0.13 (−0.23, 0.56)
Temporal Sector	0.23 (0.03, 0.41)	0.022	−0.03 (−0.23, 0.17)	0.771	0.26 (−0.17, 0.79)
Nasal Sector	0.23 (0.03, 0.41)	0.024	0.23 (0.03, 0.41)	0.025	0.00 (−0.51, 0.43)

MD = mean deviation, VF = visual field, MS = mean sensitivity, CI = 95% confidence interval.

Figure 2.

A) Scatterplot representing structure–function relationship between whole image vessel density and 10–2 visual field mean deviation change rates (r= 0.42 (0.24, 0.58), P < 0.001). B) Scatter plot representing structure–function relationship between whole image ganglion cell complex thickness change and 10–2 visual field mean deviation change rates (r= 0.27 (0.08, 0.45), P = 0.005). C) Scatter plot representing the correlation between normalized dynamic range–based slopes of vessel density loss and ganglion cell complex thinning at parafovea scan slab (R² =0.12, P < 0.001). VD = vessel density; GCC= ganglion cell complex

The SF correlation (r (95% CI)) between wiGCC and 10–2 VF MD was 0.27 (0.08, 0.45), P = 0.005 (Figure 2 B). The SF correlation (r (95% CI)) between wiGCC thinning and MS change rates of central 8 VF locations was 0.29 (0.1, 0.47), P = 0.004 and between pfGCC thinning and central 4 VF location MS change rates was 0.22 (0.04, 0.41), P = 0.026. Only superior hemiretina of wiGCC map and superior sector pfGCC map represented statistically significant SF correlation (r (95% CI)) between GCC thinning and VF MS change rates (0.43 (0.26, 0.58), P < 0.001 and 0.25 (0.05, 0.43), P = 0.011, respectively).

The correlations with LROIs of the central VF MS and regional VD and GCC maps differed by sector. There were statistically significant SF correlations at inferior and temporal locations for VD maps, which were not observed in the LROI of the GCC maps. Similarly, there was a significant correlation between the GCC superior sector and VF MS, which was not observed in the VF maps. However, there was no statistically significant difference between the SF relationships with central VF damage for VD loss and GCC thinning correlations for the whole image and parafovea maps and the corresponding LROI (all P values > 0.05 using Bootstrapping method). The average (95% CI) differences were represented in Table 3. Figure 3 represents one case of the study sample, that demonstrates progressive VD from 2015 to 2019 and also, progressive central VF defects were seen in the same period of time. However, GCC thickness is showing nearly stable thickness measurements during this follow-up period.

Figure 3.

A representing case followed up from the years 2015 to 2019, demonstrated progressive vessel density (VD) loss (top row), nearly stable ganglion cell complex (GCC) thickness (middle row) and progressive central visual field defect (bottom row).

Normalized VD change rates were faster than normalized GCC thinning over time; −7.40 (−7.71, −7.09) %/year and −2.39 (−2.94, −1.84) %/year for wiVD and wiGCC, and −6.63 (−6.96, −6.30) %/year and −2.24 (−2.79, −1.68) %/year for pfVD and pfGCC (all P < 0.001). Figure 2 C shows the weak, but significant correlation (R² =0.12, P < 0.001) between normalized VD and GCC change rates at the whole image map. The R² between normalized change rates of VD and GCC was 5% for the parafovea maps.

Discussion

In the cohort of eyes with early glaucoma damage at baseline and the average of 3.8 years of follow-up, modest, but significant longitudinal SF correlations were found between 10–2 VF measurements rates of change and both macular VD loss and macular GCC thinning. The highest longitudinal SF correlations were observed between wiVD and 10–2 VF MD. At the LROI of VD maps, there were statistically significant longitudinal SF correlations at inferior and temporal locations, which were not observed in the LROI of the GCC maps. in, general, faster-normalized change rates also were observed for global VD loss compared to global GCC thinning over time. These results support that OCTA is a promising structural modality for the detection of glaucoma progression as it is associated with central VF loss.

Although some studies have evaluated the longitudinal SF relationship between macular thickness thinning and central VF damage, there is no report regarding the relationship between longitudinal VD loss and central VF damage over time.^24,46 Such a longitudinal study is particularly important as it would more potentially reflect glaucoma progression and more efficiently estimate the structural and functional status of glaucomatous eyes. In the current study, the correlations were higher when more global parameters were investigated. This finding could be explained by reduced variability observed for global structural and functional parameters.⁴⁷ The highest longitudinal SF correlation in this study was around 0.4. Prior studies have also shown a similar longitudinal SF relationship between macular OCT and central VF.^24,46 One reason for this fairly low SF relationship is a relatively short follow-up time in the current study (mean follow-up time of 3.8 years). Moreover, since most of the eyes were receiving treatment during the follow-up period (86%), the magnitude of structural and functional change was relatively small. Finally, RGC damage might cause functional deteriorations that could be detected by functional tests with otherwise normal structural tests.^48,49

Preserving central VF for glaucoma eyes is an important task for clinicians, especially in the early stages before the eyes reach advanced stages. Compared with the standard 24–2 VF, the central VF involvement more strongly influences the quality of life of glaucoma patients and is an important predictor of glaucoma progression.^22,19 Several studies of SF relationships have focused on the correlation between different structural modalities and the central VF.^{24,46,47,50–55} In the current longitudinal study, macular OCTA VD loss and GCC thinning were both associated with central VF change. The significant correlations observed between central VF change rates and both VD and GCC change rates, globally and in most LROI, supports this finding. For SF correlation with 10–2 VF MD, VD loss was stronger than the correlation with GCC thinning, but the difference did not reach statistical significance. This is in agreement with previous cross-sectional studies that reported stronger correlations of VF measurements with VD loss compared with GCC thinning.^14,34,56,57 In the current longitudinal study, faster normalized VD loss in comparison to GCC thinning for these early glaucoma eyes was observed, which was consistent with the previous reports.¹⁰ Therefore, evaluation of VD loss could potentially estimate the overall central functional status of early-stage glaucoma eyes. However, due to the fact that faster rates of VD loss were not always associated with faster rates of central VF change (Figure 2 A), performing both OCTA and central VF would be potentially useful for monitoring glaucoma progression. Further studies are required to be designed to support this finding.

An important strength of our study is the evaluation of the SF relationship longitudinally in LROI of VD and GCC maps. Since glaucoma damage starts locally, it is important to evaluate structural and functional changes in localized areas, especially for early-stage glaucoma eyes. To address the variability of threshold sensitivity of a few central VF locations, the measurements were averaged to mean sensitivities for wiVD, pfVD and hemiretinas.⁴⁸ However, for the parafovea sectors, only one VF location was matched with the anatomical region. Another approach that was used to address the variability was performing BLUP to estimate population slope by applying random slope and random intercept effects. For the parafovea sectors, the VD loss demonstrated significant SF correlations with central VF damage observed at inferior and temporal locations, which was not observed for GCC thinning, where GCC thinning was significantly associated with central VF damage in the superior sector while VD was not. It is unclear why the sectoral associations between VF MD and VD and GCC defer. It has been reported by Hood et al. that glaucoma damage starts at an inferior–temporal location, an area known as the macular vulnerability zone.^53,55,58 The current study showed that OCTA VD loss follows the same structural trend. Regarding the hemiretinas of the whole image maps, VD loss demonstrated significant correlations with central VF damage for both superior and inferior hemiretinas. However, GCC thinning represented a significant correlation with central VF damage only in superior hemiretina. This is in contrast with a previous longitudinal SF study for macular thickness thinning and central VF which demonstrated stronger correlations in inferior hemiretina in advanced glaucoma eyes.²⁴ To evaluate glaucoma progression in inferior hemiretina for early-stage glaucoma eyes, changes in microvasculature could potentially provide additional structural information.

In this study, no statistically significant difference was found in the SF relationships between both VD loss and GCC thinning with central VF damage. This finding suggests that information from both microvasculature dropout and thickness measurements in the macula can be potentially important in monitoring glaucoma. However, a higher magnitude of SF correlations was observed between VD loss and 10–2 VF MD. Although VD had a higher correlation with central VF, the difference in the correlation of macular OCT with the central VF was not statistically significant. Future studies with larger sample sizes in different stages of glaucoma are needed to better elucidate the SF relationship of OCTA and OCT with central VF. This finding is in agreement with reports from previous studies.^16,59,60 One explanation for this finding is the theory of RGC dysfunction happens before RGC death.^49,61–65 Impaired microvasculature in the macula could result in metabolic stress insufficiency.^6,66 Therefore, VD loss could provide additional structural information which can be used to detect glaucoma progression early in the course of the disease. Additionally, it has been reported that restoration of blood supply and removing the effect of ischemia from the RGCs, could potentially reverse this RGC dysfunction and avoid RGC death.⁶⁷Thus, detection of these early defects may be important for initiating appropriate treatment to stabilize the course of glaucomatous damage.

There are several limitations to this study. First, OCTA scans with poor quality and artifacts have been excluded from the analysis. It has been shown that there is a significant number (approximately 30%) of OCTA images with artifacts.^68,69 Therefore, this analysis represents the correlations using only good-quality images. The variability introduced to the clinical dataset by these artifacts might have an effect on the SF correlation between OCTA and central VF in the actual clinical setting. Second, the incorporation of a larger 6×6 mm² could provide a more precise evaluation of SF relationships with central VF.⁷⁰ Although not shown in this study, 6×6 mm² matched with almost all the 68 locations of central VF. Future studies with an additional follow-up time of both 6×6 mm² maps and central VF could address this limitation. For the third limitation, the effect of different topical and systemic medications on the SF relationship was not considered in this study. However, some reports emphasized the effect of certain medications on vessel density loss.^71,72 Finally, for the parafovea sectors, since there was only one corresponding VF location, the variability of the VF threshold sensitivity could not be diluted by averaging the threshold sensitivities (similar to other LROIs). However, by performing BLUP for calculating the rates of change of VF measurement for those sectors, the intrinsic variability could somehow be reduced.

In conclusion, the SF correlations of VD loss and GCC thinning with central VF change rates were significant in global maps and most of the LROI. Although the correlation of VD loss over time and the rate of central function change were higher than the correlation of GCC thinning over time with central function change in this cohort of early glaucoma eyes, the magnitude of the correlations was only modest. Rates of VD loss and GCC thinning were both associated with central VF loss over time, and assessment of both should be considered for monitoring of the disease and evaluation of glaucoma progression.