Reduction of Optic Disc Microvasculature and Retinal Nerve Fiber Layer Thinning in Patients with Glaucoma

Abstract

Purpose:

To assess the relationship between the change of optic disc vessel density (ODVD) and retinal nerve fiber layer (RNFL) thinning in primary open-angle glaucoma (POAG) patients.

Design:

Retrospective case series.

Methods:

For 105 POAG patients, ≥ 5 consecutive optical coherence tomography (OCT) and OCT angiography (OCTA) images were obtained during ≥ 2 years of follow-up.

Based on enface OCTA imaging, ODVD was calculated as the ratio of pixels occupied by vessels below the internal limiting membrane within the temporal area of the optic cup, and ODVD reduction was determined when there was a statistically significant negative slope (P < 0.05) for any of the global, superior, or inferior sectors. The association between the rates of ODVD change and RNFL thinning was assessed by multivariable longitudinal linear mixed-effects model vs. time.

Results:

During 2.9 ± 0.3 years of follow-up on the 105 participants with visual field mean deviation at baseline of 5.7 ± 4.8 dB, 46 (43.8%) showed ODVD reduction. Faster global RNFL thinning was associated with the smaller Bruch’s membrane opening area (ß = 0.381; 95% confidence interval (CI), 0.120 to 0.646; P = 0.006), optic disc hemorrhage (ß = −0.567; 95% CI, −0.909 to −0.228; P = 0.002) and faster rate of global ODVD change (ß = −0.090; 95% CI, −0.139 to −0.042; P = 0.001).

Conclusions:

Reduction of optic disc microvasculature was associated with rapid RNFL thinning in POAG. This suggests a role for deep ONH circulation in the glaucoma pathogenesis.

Introduction

Glaucoma is a progressive optic neuropathy that is related to multiple factors. Although intraocular pressure (IOP) is regarded as central to its pathogenesis, there is increasing evidence that vascular factors also play a vital role.^{1, 2}

In this regard, deep-layer microvasculature dropout, as imaged by optical coherence tomography angiography (OCTA), recently has been highlighted as an important parameter related to the presence and progression of glaucoma.^3–12 It can be categorized as either optic disc microvasculature dropout (MvD-D) or parapapillary deep-layer microvasculature dropout (MvD-P), which share a common perfusion source, a short posterior ciliary artery. Despite several reports that the presence and widening of MvD-P can be a marker of rapid glaucoma progression,^8–12 little is known about the influence of MvD-D. Optic disc microvasculature directly supplies deep optic nerve head (ONH) structure, including the lamina cribrosa (LC), a key site of glaucomatous damage,^{3, 4,7,13} Therefore, change of the optic disc microvasculature can reflect alterations in ONH structure such as the development of focal LC defects, and the deformation and displacement of LC. The detection of such changes may facilitate early detection of disease progression.

This study longitudinally investigated the reduction of optic disc microvasculature detected by OCTA in eyes with primary open-angle glaucoma (POAG) and assessed its relationship with retinal nerve fiber layer (RNFL) thinning rate.

Methods

The present study was approved by the Institutional Review Board of Haeundae Paik Hospital. We reviewed the medical records of consecutive POAG patients followed by a glaucoma specialist (M.H.S) between October 2019 and September 2023. Informed consent was waived due to the retrospective nature of this study.

Participants

Initial ophthalmologic examinations included measurement of best-corrected visual acuity (BCVA), refraction, slit-lamp biomicroscopy, IOP by Goldmann applanation tonometry, gonioscopy, central corneal thickness (CCT) measured by the Pentacam Scheimpflug imaging system (Oculus Optikgeräte GmbH, Germany), axial length (AXL) by IOL Master (Carl Zeiss Meditec, Dublin, CA, USA), dilated stereoscopic examination of the optic disc, simultaneous color and red-free fundus photography (TRC-NW8; Topcon, Tokyo, Japan), standard automated perimetry (Humphrey Field Analyzer; 30–2 Swedish Interactive Threshold Algorithm; Carl-Zeiss Meditec).

A minimum of 5 good-quality spectral-domain optical coherence tomography (SD-OCT) (Spectralis; Heidelberg Engineering GmbH, Germany) and swept-source (SS)-OCTA (Angioplex; Carl Zeiss Meditec, Germany) images were obtained for ≥ 2 years at 4–6-month intervals. Each OCT and OCTA imaging sessions were performed within 4 months. If both eyes were eligible for inclusion in the study, one eye was randomly selected.

Subjects were required to have been diagnosed with POAG, to have best corrected visual acuity of 20/30, and to be older than 18 years. Those with a history of ocular surgery, including cataract or glaucoma surgery, and other intraocular or neurologic diseases (e.g., pituitary tumor) that could cause visual field (VF) loss were excluded. Those with diabetes mellitus and/or systemic hypertension were included unless they had been diagnosed with diabetic or hypertensive retinopathy.

POAG was defined as glaucomatous optic nerve damage (i.e., focal thinning, notching, localized or diffuse atrophy of the RNFL) with compatible VF defect and an open angle by gonioscopy. Glaucomatous VF damage was defined as outside the normal limits on a glaucoma hemifield test, or 3 abnormal points with P<5% probability of being normal, one with P<1% by pattern deviation, or pattern standard deviation (PSD) of P <5% on 2 consecutive, reliable (≤33% fixation losses and false negatives, ≤15% false positives) tests.^8,14 Baseline IOP was measured prior to initiating IOP-lowering treatment or as indicated in the referral notes.⁹ The mean follow-up IOP and IOP fluctuation were calculated as the average and standard deviation of all IOPs during the entire follow-up period. The mean follow-up IOP was calculated as the average IOP for every 6 months.^{8, 9} Optic disc hemorrhage (DH), defined as an isolated splinter- or flame-shaped hemorrhage on the ONH, had to be detected at least once during the follow-up period based on fundus photographs and clinical examinations performed at intervals of 3 to 6 months. ^{8, 11,15–19}

Analysis of Focal LC Defect and Parapapillary ß-zone on SD-OCT

Based on 48 enhanced depth imaging radial B-scans providing a 20° × 20° high-resolution scan pattern of Spectralis OCT2, parapapillary ß-zone (ß-zone) and focal LC defect were determined. ß-zone was defined as an area lacking retinal pigment epithelium (RPE) with a temporal width ≥ 200μm on at least 1 radial scan,^20,21 and focal LC defect as laminar holes or disinsertion violating the normal U- or W-shaped contour of the anterior laminar surface of which the diameter was at least 100 mm.^22,23,24,25 Presence of DH, ß-zone, and focal LC defect was determined by 2 masked reviewers (M.H.S and Y.J.L), and disagreements were resolved by consensus. If consensus could not be reached, the subjects were excluded.

Rate of Circumpapillary RNFL Thinning on SD-OCT

Circumpapillary RNFL thickness in a global area and 6 sectors (superotemporal (TS), inferotemporal (TI), temporal (T), superonasal (NS), inferonasal (NI), nasal (N)) was derived at each point along the circumference of a 3.5 mm diameter circle using Spectralis OCT2 Glaucoma Module Premium Edition software. The accuracy of the RNFL segmentation was reviewed, and segmentation errors were corrected manually by one masked reviewer (J.W.S).⁹ Follow-up OCT imaging was performed using a built-in automated realignment procedure (referred to as the follow-up examination in the system documentation).^9,12 The rate of RNFL thinning was derived on a per-year basis.

Reduction of Optic Disc and Parapapillary Deep-Layer Vessel Density

Angioplex OCTA incorporated into the PLEX Elite 9000 SS-OCT system provided 6.0ⅹ6.0 mm ONH images.^4,26 The SS-OCTA system utilized FastTrac motion tracking and projection artifact removal techniques to minimize possible motion and projection artifacts during imaging.^4,26,27 Poor-quality OCTA images (e.g., poor clarity, signal strength < 7, motion artifact, segmentation error, poor visualization of deep-layer microvasculature (i.e., limited penetration of OCT beam, shadowing of neuroretinal rim or superficial vessels)) were excluded.⁴

For measurement of optic disc vessel density (ODVD), a whole-signal-mode image below the internal limiting membrane was derived;^{4, 28} for parapapillary deep-layer vessel density (PDVD), slabs below the Bruch’s membrane (BM) containing microvasculature within the choroid or the scleral flange were used.^{5, 14} En-face OCTA images were binarized using Otsu’s method, which utilizes publicly available ImageJ software (National Institutes of Health, Bethesda, MD, USA; http://imagej.nih.gov/ij/ ) under the assumption that images contain two classes of pixels following a bimodal distribution.^27,28 This method searches the optimum threshold that minimizes intraclass variance while maximizing interclass variance. ODVD and PDVD were calculated as the ratio of black pixels occupied by vessels divided by those from the region of interest (ROI). With reference to a confocal scanning laser ophthalmoscopy (CSLO) image, the ROIs of ODVD and PDVD were derived as the temporal area of the vertical line passing through the optic disc centroid, since the nasal side is hardly assessable due to shadowing of the large vessels and neural rim (Figure 1).^{4, 6} Similarly, the ROI of the ODVD also was determined as the cup area after excluding the neuroretinal rim.^28,29 The ROI of the PDVD was defined as the ß-zone area, and, in eyes without ß-zone, as the parapapillary area within 200 μm of the optic disc boundary (Figure 1). This was done because MvD-P can be observed in an area without ß-zone.²¹ The global area and the superior and inferior hemispheres were derived according to the fovea-Bruch’s membrane opening (BMO) axis (Figure 1). A linear regression analysis vs. time was applied to the global and hemispheric ODVD and PDVD for each subject to detect the rate of change (expressed in % per year) of MvD-D and MvD-P over time. The average of measurements by two independent observers (M.H.S and H.J.K) masked to the clinical characteristics of the subjects were calculated. Reduction of ODVD was determined when there was a statistically significant negative slope (P < 0.05) of ODVD for any of the global, superior, or inferior areas (Figure 1, 2).⁸.³⁰ Disagreements between the two observers were resolved by consensus, and if consensus could not be reached, the subjects were excluded.

FIGURE 1.

Measurement of optic disc vessel density (ODVD) and parapapillary deep vessel density (PDVD) based on OCTA. CSLO (A), enface OCTA image at baseline (B), follow-up OCTA images (C), scatterplot showing rate of ODVD change in inferior hemisphere (D), serial maps of sectoral RNFL thickness (E), and scatterplot showing rate of RNFL thinning of TI sector (F). B2 and B3 are the same as B1, indicating the optic disc margin (red dotted circle) and temporal half of the optic disc cup for measurement of the ODVD (blue semicircle), a parapapillary area within 200 μm of the optic disc margin for measurement of the PDVD in eyes without parapapillary ß-zone (red solid line), the fovea-Bruch’s membrane opening axis for division of the superior and inferior hemispheres (blue dotted line) (B2), and the relatively preserved optic disc microvasculature in the inferior area (green arrows, B3). Note that notable deterioration of ODVD (red arrows, C and D) preceded that of RNFL thinning (blue arrows, E and F) and that there was no remarkable change in PDVD.

FIGURE 2.

Representative cases showing differing rates of sectoral retinal nerve fiber layer (RNFL) thinning between eyes with (A, B) and without (C, D) optic disc vessel density (ODVD) reduction. Serial whole-signal-mode en-face optical coherence tomography angiography (OCTA) images (A1, C1), serial maps of sectoral RNFL thickness (B1, D1), and scatterplots demonstrating the rate of ODVD change (A2, C2) and that of sectoral RNFL thinning (B2, D2). A and B, An eye with ODVD reduction had rapid deterioration of ODVD in the superior hemisphere (−8.09%/yr, P = 0.017; A1, A2) and rapid RNFL thinning of the compatible TS sector (−5.04 μm/yr, P = 0.019; B1, B2). C and D, An eye with stable ODVD in the inferior hemisphere (0.05%/yr, P = 0.989; C1, C2) had stable RNFL in the compatible TI sector (0.36 μm/yr, P = 0.270; D1, D2). Note that there was no notable change in the parapapillary deep-layer microvasculature in either eye.

Statistical Analysis

Baseline characteristics and test results were compared between the eyes with and without ODVD reduction using the Independent t-test. Categorical variables were compared using Fisher’s exact test or Chi-squared test_.³¹ To identify factors associated with the rate of global RNFL thickness and ODVD change, longitudinal linear mixed-effects modeling was performed.^32,33 This model predicted longitudinal global RNFL thickness and ODVD measurements across follow-up time, fit with an interaction term between follow-up time and independent covariates of interest. All linear mixed-effects models were fit with a random intercept to account for within-subject correlation, a random slope across follow-up time, as well as auto-correlation components, to describe within-group correlation structures, and fixed variance weights to describe within-group heteroscedasticity structures.

Logistic regression analysis was performed to identify factors associated with ODVD reduction. Variables with a value < 0.1 in the univariable regression analysis were included in the multivariable model. Kappa values were used to measure the inter-observer agreement for determination of focal LC defect, ß-zone, DH, and ODVD reduction. The intraclass coefficient (ICC) was calculated to assess the inter-observer reproducibility for measurement of baseline values and the rates of change of global ODVD and PDVD. The analyses were performed using MedCalc (MedCalc, Inc., Mariakerke, Belgium) and R software version 4.2.3 (R Project for Statistical Computing, Vienna, Austria), and P values < 0.05 were considered statistically significant.

Results

Among the 132 eyes with POAG that met the inclusion and exclusion criteria, 27 eyes (20.5%) with poor-quality OCTA images (n = 19) or failure to reach consensus in determining reduction of ODVD and/or PDVD (n = 8) were excluded, thus leaving 105 eyes for analysis. There were excellent inter-observer agreements for determination of focal LC defect (Kappa = 0.86), ß-zone (Kappa = 0.88), DH (Kappa = 0.85), and ODVD reduction (Kappa = 0.83).

There was good inter-observer reproducibility for the ODVD measurements at baseline (ICC = 0.89 for the global and inferior areas, and 0.90 for the superior area), PDVD at baseline (ICC = 0.88, 0.89, and 0.91 for global, superior and inferior areas), rate of ODVD change (ICC = 0.88 for all three areas), and rate of PDVD change (ICC =0.88, 0.83, and 0.87 for global, superior and inferior areas).

Among the 105 eyes included in the study, ODVD reduction was observed in 46 eyes (43.8%) and not in 59 eyes (56.2%). Among the 46 eyes showing reduction of ODVD, in 10 eyes (21.7%) reduction was observed in the superior area, and 18 (39.1%) eyes in the inferior and 18 (39.1%) in both areas, respectively.

Table 1 compares the clinical features of the eyes with and without ODVD reduction. Age, gender, AXL, CCT, number and types of topical medications, BMO area, presence of ß-zone, follow-up period, mean number of OCTA images, baseline IOP, RNFL thicknesses except NS sector, mean deviation (MD), PSD, PDVD, and ODVD at baseline did not differ between the two groups (P > 0.1 for all comparisons). Eyes with ODVD reduction had a significantly larger number of OCT images, higher IOP fluctuation during follow-up, thicker baseline RNFL of NS sector, and higher prevalence of DH (P < 0.05 for all comparisons). Eyes with ODVD reduction had a significantly faster rate of RNFL thinning than those without reduction in all areas (P < 0.05) except the NS (P = 0.055) and N sectors (P = 0.163) (Table 1, Figure 1, 2). They also showed faster rates of ODVD and PDVD change in all areas (P < 0.05) except the superior hemisphere for PDVD change (P = 0.133) (Table 1).

Table 1.

Comparison of clinical characteristics between glaucomatous eyes with and without optic disc vessel density (ODVD) reduction

Variables	ODVD reduction (n = 46)	Without ODVD reduction (n = 59)	P-value
Age (years) (range)	53.0 ± 13.3 (27 to 80)	57.2 ± 12.7 (28 to 81)	0.105*
Gender (male/female)	18/28	18 /41	0.358†
Axial length (mm)	25.3 ± 1.9	24.9 ± 1.4	0.195*
Central Corneal Thickness (μm)	538.4 ± 35.9	528.2 ± 35.4	0.151*
Self-reported diabetes, n (%)	3 (6.5 %)	4 (6.8 %)	1.000 ‡
Self-reported hypertension, n (%)	10 (21.7 %)	7 (11.9 %)	0.175†
Number of topical glaucoma medications			0.313‡
0	1	6
1	9	11
>1	36	42
Topical medications, n			0.781†
Prostaglandin analogs	36	38
Beta-antagonists	34	42
Carbonic anhydrase inhibitors	33	36
Alpha-1 agonist	24	20
BMO area (mm2)	2.41 ± 0.63	2.36 ± 0.59	0.656*
Presence of ßPPA (%)	41 (89.1%)	48 (81.4 %)	0.274‡
Follow-up period (yrs)	2.9 ± 0.3	2.9 ± 0.3	0.429*
Number of OCT images (range)	5.6 ± 1.0 (5 to 9)	5.3 ± 0.5 (5 to 7)	0.021 *
Number of OCTA images (range)	5.2 ± 0.6 (5 to 8)	5.1 ± 0.4 (5 to 7)	0.326*
IOP (mmHg)
Baseline	18.2 ± 4.7	16.9 ± 4.5	0.151*
Average	12.6 ± 2.0	11.9 ± 1.6	0.085*
Fluctuation	1.5 ± 0.6	1.2 ± 0.5	0.014 *
RNFL thickness at baseline (μm)
Global	75.8 ±13.6	72.9 ±15.6	0.316*
TS	106.6 ± 36.6	96.3 ± 36.7	0.155*
TI	67.9 ± 34.3	74.7 ± 35.9	0.329*
NS	104.7 ± 23.2	92.7 ± 30.1	0.028 *
NI	82.2 ± 26.2	81.3 ± 25.3	0.865*
T	61.0 ± 17.6	61.9 ± 14.2	0.772*
N	66.9 ± 16.9	63.6 ± 16.8	0.320*
VF MD at baseline (dB)	−5.7 ± 4.2	−5.6 ± 5.3	0.974*
VF PSD at baseline (dB)	7.2 ± 4.4	6.0 ± 4.6	0.195*
Detection of DH during follow-up, n (%)	21 (45.7 %)	16 (27.1 %)	0.049 †
Presence of focal LC defect during follow-up, n (%)	33 (71.7 %)	33 (55.9 %)	0.098†
PDVD at baseline (%)
Total	87.8 ± 5.0	87.4 ± 5.3	0.689*
Superior	90.5 ± 5.8	90.3 ± 4.8	0.848*
Inferior	85.3 ± 6.6	84.1 ± 8.3	0.449*
ODVD at baseline (%)
Total	65.4 ± 9.6	63.5 ± 12.3	0.396*
Superior	67.7 ± 11.8	67.5 ± 13.1	0.921*
Inferior	64.7 ± 10.6	61.9 ± 12.8	0.233*
Rate of RNFL thinning (μm/yr)
Global	−1.5 ± 1.3	−0.6 ± 0.7	< 0.001 *
TS	−2.6 ± 3.3	−1.0 ± 1.7	0.002 *
TI	−2.8 ± 3.7	−0.8 ± 1.4	<0.001 *
NS	−1.4 ± 2.2	−0.7 ± 1.7	0.055*
NI	−1.4±2.2	−0.6 ± 1.5	0.048 *
T	−1.3±1.3	−0.6±1.0	0.002 *
N	−0.7±1.9	−0.3±0.9	0.163*
Rate of ODVD change (%/yr)
Global	−6.3±3.1	−1.4±2.2	< 0.001 *
Superior	−7.2±3.6	−1.6±2.4	< 0.001 *
Inferior	−6.4±3.4	−1.0±2.3	< 0.001 *
Rate of PDVD change (%/yr)
Global	−1.2±1.9	−0.3±1.1	0.001 *
Superior	−0.6±1.4	−0.2±1.1	0.133*
Inferior	−1.5±2.1	−0.2±1.6	0.001 *

Values are mean ± standard deviation unless otherwise indicated.

^*The comparison was performed by using an Independent sample t-test.

^†The comparison was performed by using Chi-square test.

^‡The comparison was performed by using Fisher’s exact test.

P values < 0.05 are noted in boldface.

BMO = Bruch’s membrane opening; ßPPA = parapapillary ß-zone; OCT = optical coherence tomography; OCTA = optical coherence tomography angiography; IOP = intraocular pressure; RNFL = retinal nerve fiber layer; TS = superotemporal; TI = inferotemporal; NS = superonasal; NI = inferonasal; T = temporal; N = nasal; VF = visual field; MD = mean deviation; PSD = pattern standard deviation; dB = decibel; DH = optic disc hemorrhage; LC = lamina cribrosa; PDVD = parapapillary deep-layer vessel density.

Table 2 provides the results of the longitudinal linear mixed-effects model assessing factors associated with rate of global RNFL thinning. In the univariable analysis, smaller BMO area, (ß = 0.310; 95% confidence interval (CI), 0.006 to 0.612) presence of DH (ß = −0.643; 95% CI, −1.019 to −0.267; P = 0.001) and focal LC defect (ß = −0.469; 95% CI, −0.852 to −0.086; P = 0.018) during follow-up, faster rates of global PDVD change (ß = −0.131; 95% CI, −0.253 to −0.009; P = 0.037) and global ODVD change (ß = −0.103; 95% CI, −0.151 to −0.054; P < 0.001) were associated with faster rate of global RNFL thinning (Table 2). In the multivariable analysis, smaller BMO area (ß = 0.381; 95% CI, 0.120 to 0.646; P = 0.006), presence of DH (ß = −0.567; 95% CI, −0.909 to −0.228; P = 0.002) and faster rate of global ODVD change (ß = −0.090; 95% CI, −0.139 to −0.042; P = 0.001) remained as significant factors associated with the rate of global RNFL thinning (Table 2, Figure 1, 2). In the assessment of factors related to the rate of global ODVD change (Table 3), male gender (ß = −1.497; 95% CI, −2.913 to −0.078; P = 0.041), higher baseline IOP (ß = −0.156; 95% CI, −0.304 to −0.008; P = 0.040), larger number of OCT images (ß = −1.212; 95% CI, −2.074 to −0.350; P = 0.007), higher global ODVD at baseline (ß = −0.141; 95% CI, −0.199 to −0.083; P < 0.001), faster rate of global PDVD change (ß = −0.958; 95% CI, −1.373 to −0.543; P <0.001), and faster rate of global RNFL thinning (ß = −0.937; 95% CI, −1.598 to −0.276; P = 0.007) were significantly associated in the univariable analysis (Table 3). In the multivariable analysis, larger number of OCT images (ß = −0.844; 95% CI, −1.606 to −0.086; P = 0.036), higher global ODVD at baseline (ß = −0.120; 95% CI, −0.172 to −0.067; P < 0.001), faster rate of global PDVD change (ß = −0.635; 95% CI, −1.024 to −0.242; P = 0.002), and faster rate of global RNFL thinning (ß = −0.631; 95% CI, −1.214 to −0.045; P = 0.042) remained as significant factors associated with the rate of global ODVD change (Table 3, Figure 1, 2). When an additional analysis was performed for assessment of factors associated with ODVD reduction (Table 4), it was found that IOP fluctuation (OR: 2.579 per 1 mmHg higher; P = 0.043), faster rate of global PDVD change (OR: 1.595 per 1%/yr faster; P = 0.020), and faster rate of global RNFL thinning (OR: 2.488 per 1 μm/yr faster; P = 0.003) were associated with ODVD reduction. Average IOP (P = 0.089), presence of DH (P = 0.051), and focal LC defect (P = 0.099) were associated with marginal significance in the univariable model but not in the multivariable model (P > 0.10).

Table 2.

Factors associated with the rate of global retinal nerve fiber layer (RNFL) thinning

Variables	Univariable Model		Multivariable Model
	Estimate (95% CI)	p-value	Estimate (95% CI)	p-value
Age at baseline, per 1 yr older	−0.007 (−0.022, 0.007)	0.325
Sex, male	−0.053 (−0.449, 0.342)	0.793
CCT at baseline, per 1 μm thinner	0.004 (−0.001, 0.009)	0.133
Axial length at baseline, per 1 mm longer	0.042 (−0.0 71, 0.154)	0.468
Baseline IOP, per 1 mmHg higher	−0.013 (−0.055, 0.028)	0.531
Average IOP, per 1 mmHg higher	−0.085 (−0.190, 0.019)	0.113
IOP fluctuation, per 1 mmHg higher	−0.146 (−0.468, 0.175)	0.375
Total follow-up period, per 1 yr longer	0.199 (−0.382, 0.779)	0.504
Number of OCT images, per 1 more	−0.018 (−0.264, 0.227)	0.883
Global RNFL thickness at baseline, per 1 μm thinner	0.007 (−0.006, 0.020)	0.275
VF MD at baseline, per 1 dB worse	−0.004 (−0.043, 0.035)	0.845
PPPA, presence	0.800 (−0.116, 1.717)	0.090	0.600 (−0.219, 1.413)	0.160
BMO area, per 1 mm2 larger	0.310 (0.006, 0.612)	0.048	0.381 (0.120, 0.646)	0.006
Detection of DH during follow-up period	−0.643 (−1.019, −0.267)	0.001	−0.567 (−0.909, −0.228)	0.002
Detection of focal LC defect during follow-up period	−0.469 (−0.852, −0.086)	0.018	−0.308 (−0.641, 0.031)	0.081
Global PDVD at baseline, per 1% higher	−0.020 (−0.057, 0.017)	0.285
Global ODVD at baseline, per 1% higher	0.001 (−0.016, 0.018)	0.901
Rate of the Global PDVD change, per 1 %/yr faster	−0.131 (−0.253, −0,009)	0.037	−0.030 (−0.149, 0.089)	0.627
Rate of the Global ODVD change, per 1 %/yr faster	−0.103 (−0.151, −0.054)	< 0.001	−0.090 (−0.139, −0.042)	0.001

^*Adjusted for all variables with P < 0.10 in the univariable model

P values < 0.05 are noted in boldface

CCT = central corneal thickness; IOP = intraocular pressure; OCT = optical coherence tomography; VF = visual field; MD = mean deviation; dB = decibel; ßPPA = ß-zone parapapillary atrophy; BMO = Bruch’s membrane opening; DH = optic disc hemorrhage; LC = lamina cribrosa; PDVD = parapapillary deep-layer vessel density; ODVD = optic disc vessel density.

Table 3.

Factors associated with the rate of global optic disc vessel density (ODVD) change

Variables	Univariable Model		Multivariable Model
	Estimate (95% CI)	p-value	Estimate (95% CI)	p-value
Age at baseline, per 1 yr older	0.031 (−0.021, 0.084)	0.245
Sex, male	−1.497 (−2.913, −0.078)	0.041	−1.010 (−2.209, 0.195)	0.112
CCT at baseline, per 1 μm thinner	0.014 (−0.005, 0.033)	0.152
Axial length at baseline, per 1 mm longer	0.220 (−0.632, 0.191)	0.297
Baseline IOP, per 1 mmHg higher	−0.156 (−0.304, −0.008)	0.040	−0.141 (−0.288, 0.006)	0.068
Average IOP, per 1 mmHg higher	−0.370 (−0.748, 0.008)	0.058	0.057(−0.318, 0.431)	0.770
IOP fluctuation, per 1 mmHg higher	−0.978 (−2.146, 0.192)	0.104
Total follow-up period, per 1 yr longer	0.329 (−1.797, 2.458)	0.762
Number of OCT images, per 1 more	−1.212 (−2.074, −0.350)	0.007	−0.844 (−1.606, −0.086)	0.036
Global RNFL thickness at baseline, per 1 μm thinner	0.028 (−0.019, 0.075)	0.244
VF MD at baseline, per 1 dB worse	0.072 (−0.071, 0.215)	0.328
ßPPA, presence	2.430 (−0.929, 5.792)	0.159
BMO area, per 1 mm2 larger	−0.425 (−1.563, 0.713)	0.465
Detection of DH during follow-up period	−0.595 (−2.035, 0.847)	0.420
Detection of focal LC defect during follow-up period	−0.601 (−2.028, 0.825)	0.411
Global PDVD at baseline, per 1% higher Global ODVD at baseline, per 1% higher	−0.107 (−0.238, 0.026)	0.117
	−0.141 (−0.199, −0.083,)	< 0.001	−0.120 (−0.172, −0.067,)	< 0.001
Rate of the Global PDVD change, per 1 %/yr faster	−0.958 (−1.373, −0.543)	< 0.001	−0.635 (−1.024, −0.242)	0.002
Rate of the Global RNFL thinning, per 1 μm/yr faster	−0.937 (−1.598, −0.276)	0.007	−0.631 (−1.214, −0.045)	0.042

^*Adjusted for all variables with P < 0.10 in the univariable model.

P values < 0.05 are noted in boldface.

CCT = central corneal thickness; IOP = intraocular pressure; OCT = optical coherence tomography; RNFL = retinal nerve fiber layer; VF = visual field; MD = mean deviation; dB = decibel; ßPPA = ß-zone parapapillary atrophy; BMO = Bruch’s membrane opening; DH = optic disc hemorrhage; LC = lamina cribrosa; PDVD = parapapillary deep-layer vessel density.

Table 4.

Factors associated with the reduction of optic disc vessel density (ODVD).

Variables	Univariable Model		Multivariable Model*
	Odds ratio, 95 % CI	P value	Odds ratio, 95 % CI	P value
Age at baseline, per 1 yr older	0.975 (0.946 – 1.005)	0.107
Male	1.464 (0.651 – 3.295)	0.357
CCT at baseline, per 1μm thinner	0.992 (0.981 – 1.003)	0.152
Axial length at baseline, per 1mm longer	1.169 (0.924 – 1.478)	0.194
Baseline IOP, per1mmHg higher	1.065 (0.976 – 1.163)	0.157
Average IOP, per1mmHg higher	1.213 (0.971 – 1.154)	0.089	1.020 (0.756 – 1.376)	0.898
IOP fluctuation, per1mmHg higher	2.379 (1.152 – 4.914)	0.019	2.579 (1.032 – 6.444)	0.043
Total follow-up period, per 1 yr longer	1.628 (0.491 – 5.393)	0.425
Number of OCTA tests, per 1 more	1.446 (0.687 – 3.042)	0.331
Global RNFL thickness at baseline, per 1μm thicker	1.014 (0.987 – 1.041)	0.314
VF MD at baseline, per 1 dB worse	1.001 (0.928 – 1.080)	0.976
ßPPA, presence	1.880 (0.603 – 5.854)	0.277
BMO area, per 1 mm2 larger	1.158 (0.612 – 2.192)	0.653
Detection of DH during follow-up period	2.258 (0.998 – 5.106)	0.051	1.269 (0.455 – 3.543)	0.649
Detection of focal LC defect during follow-up period	2.000 (0.879 – 4.551)	0.099	1.335 (0.496 – 3.588)	0.568
Global PDVD at baseline, per 1%	0.985 (0.916 – 1.061)	0.686
worse
Global ODVD at baseline, per 1% worse	0.985 (0.951 – 1.020)	0.393
Rate of the global PDVD progression, per 1 %/yr faster	1.684 (1.175 – 2.412)	0.005	1.595 (1.075 – 2.366)	0.020
Rate of global RNFL thinning, per 1 μm/yr faster	2.934 (1.676 – 5.139)	<0.001	2.488 (1.352 – 4.575)	0.003

^*Adjusted for all variables with P < 0.10 in the univariable Model.

P values < 0.05 are noted in boldface.

CCT = central corneal thickness; IOP = intraocular pressure; OCTA = optical coherence tomography angiography; RNFL = retinal nerve fiber layer; VF = visual field; MD = mean deviation; dB = decibel; ßPPA = parapapillary ß-zone; BMO = Bruch’s membrane opening; DH = optic disc hemorrhage; LC = lamina cribrosa; PDVD = parapapillary deep-layer vessel density.

The correlation of the global rates of change between RNFL, ODVD, and PDVD was demonstrated in Figure 3. There was a statistically significant positive relationship for all comparisons (r = 0.29; P = 0.003 for rates of RNFL vs. ODVD slopes, r = 0.42; P < 0.001 for PDVD vs. ODVD slopes, and r = 0.22; P = 0.027 for RNFL vs. PDVD slopes).

FIGURE 3.

The correlation of the global rates of change of retinal nerve fiber layer (RNFL) vs. optic disc vessel density (ODVD) (A), those of parapapillary deep-layer vessel density (PDVD) and ODVD (B), and of RNFL and PDVD (C).

To assess the topographic relationship between ODVD reduction and RNFL thinning, we plotted the rate of ODVD and RNFL changes sectorally (Figure 4). There was a positive relationship between the rates of ODVD and RNFL change in both the superior (r =0.28; P = 0.004) and inferior (r =0.21; P = 0.035) areas. In a subgroup analysis comparing the RNFL thinning rate between eyes with ODVD reduction only in the inferior area (n = 18) and those without ODVD reduction in all areas (n = 59) (Figure 5), significantly faster rate of RNFL thinning of those with ODVD reduction was observed only in the T (−2.0 ± 1.8 vs. −0.6 ± 1.0 μm/year; P = 0.001) and TI (−4.3 ± 4.8 vs. −0.8 ± 1.4 μm/year; P < 0.001) sectors. Given that the number of eyes for which ODVD reduction was observed only in the superior hemisphere was too small for meaningful statistical analysis (n = 10), subgroup analysis on these subsets was not performed.

FIGURE 4.

Association between rates of optic disc vessel density (ODVD) and retinal nerve fiber layer (RNFL) changes in superior (A), and inferior (B) areas.

FIGURE 5.

Comparison of retinal nerve fiber layer (RNFL) thinning rate between eyes with optic disc vessel density (ODVD) reduction only in the inferior area (n = 18) and those without reduction in all areas (n = 59). Significantly faster rate of RNFL thinning of those with ODVD reduction was observed only in the TI (P < 0.001) and T (P = 0.001) sectors.

Discussion

In this longitudinal study, reduction of the optic disc microvasculature was observed in 46 eyes (43.8%) of POAG patients. This was independently associated with faster RNFL thinning, even after adjusting for known influencing factors such as focal LC defect, DH, and change of MvD-P. Moreover, rapid RNFL thinning was observed in a corresponding area with ODVD reduction, providing evidence of a topographic relationship between ODVD reduction and RNFL thinning. These findings suggest that optic disc microvasculature deterioration can be a useful marker of glaucoma progression.

The present study concurs with previous cross-sectional studies on the association of MvD-D with glaucomatous optic disc damage.^{3, 4, 6} Furthermore, we found that ODVD reduction is closely related to disease progression. However, it is unclear whether impaired deep ONH circulation plays a causative role in the pathogenesis of glaucoma or is simply an epiphenomenon. Given that the LC is the primary site of glaucomatous damage, reduced optic disc perfusion can be an early sign of disease progression.^3,6 In this study, several eyes showed notable ODVD reduction before subsequent RNFL thinning (Figure 1, 2). This finding suggests that ODVD reduction can be a sensitive marker that precedes RNFL thinning. However, prospective longitudinal studies are needed to clarify the temporal relationship between the vascular and structural parameters of the deep ONH.

Meanwhile, faster rate of ODVD change was associated with higher ODVD at baseline. Given ODVD’s positive association with disease severity,^{3, 6} optic disc microvasculature changes may help detect progression in early glaucoma.

The rate of PDVD change was associated with faster RNFL thinning in the univariable regression but not in the multivariable model. This result does not concur with previous reports that enlargement of MvD-P is related to glaucoma progression.^{9, 10, 12} A possible explanation is that optic disc circulation directly supplies the LC, a primary site of damage, while parapapillary deep-layer microvasculature is located adjacent to the optic disc. Therefore, impaired optic disc circulation might reflect disease progression better than impaired parapapillary circulation. Also, the rates of change of parapapillary and optic disc microvasculature were significantly correlated in this study (ß = 0.300; P < 0.001). Second, 15.2% (16/105) of the study population lacked β-zone, and all of them showed stable PDVD, while 31.3% (5/16) showed a change of ODVD (P = 0.043). Although MvD-P can be observed in several eyes without β-zone,²¹ it is difficult to detect its changes due to the signal attenuation of the pigmented RPE.³⁴

The current results indicating that DH was associated with RNFL thinning add to the literature showing a significant relationship between DH and glaucoma progression.^{11, 15–19, 35} With ODVD reduction, DH was significantly associated in the univariable model, but not in the multivariable regression. In light of the previous reports of a relationship between DH and MvD-P presence and enlargement,¹¹ further investigations are needed to elucidate its relationship with ODVD reduction.

Focal LC defect is known to be one of the essential factors associated with glaucoma progression.^{22, 36} In this study, focal LC defect was related to RNFL thinning marginally but became statistically significant after excluding ODVD from the multivariable model (data not shown). Therefore, the current result does not necessarily contradict the influence of focal LC defect on glaucoma progression, though it does suggest that it has less influence than that of MvD-D. Meanwhile, no significant relationship existed between focal LC defect and ODVD change. This finding is intriguing in that, similar to previous studies,^{4, 6} eyes with focal LC defect had significantly lower ODVD at baseline than those without focal LC defect (P = 0.018). Prospective longitudinal studies with extended follow-up periods will clarify the temporal relationship between enlargement of focal LC defect and ODVD reduction.

According to the logistic regression analysis with the exception of a statistically significant association between the IOP fluctuation and ODVD reduction, IOP parameters were not associated with either RNFL thinning or ODVD reduction. Possible reasons are the exclusion of patients who had undergone glaucoma surgery to minimize the effect of sudden IOP change on ocular circulation. Also, the study population was comprised mainly of subjects with IOP < 22 mmHg (85% (89/105)). Meanwhile, there was no significant difference in the number and types of glaucoma medications between eyes with and without ODVD reduction (Table 1), suggesting that the influence of glaucoma medication on the relationship between the ODVD reduction and RNFL thinning is not substantial. Further studies that include subjects representative of a wider range of IOP and that assess the influence of treatment are warranted.

Previous studies defined the progression of MvD-P as widening of its circumferential extent.^9,10,12 Despite the clinical relevance, this method is limited by subjectivity and arbitrary criteria because observers need to delineate the MvD-P which is often scattered in multiple areas. Also, this method is often not useful for identifying newly developed MvD-P in eyes without MvD-P at baseline.⁹ To overcome these shortcomings, in this study, the density of the vasculature was calculated on a sectoral basis. Also, we defined the “reduction of ODVD and PDVD” as binary yes or no variables based on whether there was significant negative slope (P <0.05) (Table 4). In contrast, we used a continuous rate of change of these variables in Table 3 to investigate factors affecting how fast the eyes were progressing. These are two different ways to characterize progression, and both are important. For example, several eyes have large negative slopes that do not reach statistical significance (eg. slope: −3.2 %/yr P = 0.24) mainly because of high test-retest variability, and vice versa (eg. slope: −1.5 %/yr P =0.01).

This study had several limitations. In our trend-based quantitative analysis of ODVD, only the cup area was included, due to shadowing of the neuroretinal rim. Therefore, subtle ODVD changes under the rim area may have been overlooked.^{26, 37} The clinical utility of OCTA for detection of progression still needs to be clarified, owing to its high test-retest variability and lack of gold standard. However, there was good inter-observer agreement (ICC range: 0.83 – 0.91, Kappa = 0.86). Second, this study may be limited by the study population that was myopic and had low untreated IOP on average, retrospective nature, referral glaucoma clinic and relatively short follow-up period. Therefore, the current result may not be applied to the general population. In addition, we did not utilize VF measurements which often serve as the gold standard for defining progression since there often is a longer time needed to detect changes than with the RNFL. However, although one eye was randomly selected if both eyes were eligible for the study, ODVD reduction was observed in a considerable number of subjects (43.8% (46/105)) in a short follow-up time; this suggests that the follow-up period was long enough to detect it. Including visual fields to define progression is planned for a future project when longer follow-up is available. Finally, subjects with an optic cup area large enough to permit measurement of the ODVD were included in this study; thus, there may have been selection bias incurred by excluding those with a small optic disc and/or thick neuroretinal rim. Also, a fairly large number of subjects (20.5%) either had unusable scans or images that could not be adjudicated. Further improvement of the OCTA technique is needed.

In conclusion, deterioration of optic disc microvasculature as detected by OCTA was an independent factor associated with RNFL thinning, even after adjusting for other influencing factors such as IOP, DH, focal LC defect, and MvD-P. Moreover, ODVD reduction and the RNFL thinning were spatially correlated. These results suggest that monitoring OCTA-derived optic disc microvasculature changes over time can provide an additional strategy for early detection of progression. Future prospective longitudinal studies may help to clarify the role of deep optic disc circulation in the pathogenesis of glaucoma.

Table of Contents

During a 2.9-year follow-up on 105 subjects, 46 (43.8%) showed reduction of optic disc vessel density (ODVD). A faster rate of ODVD reduction was an independent factor for rapid retinal nerve fiber layer (RNFL) thinning, even after adjusting for other influencing factors such as disc hemorrhage (DH), focal lamina cribrosa (LC) defect, and parapapillary deep-layer microvasculature density. This finding suggests that reduction of optic disc microvasculature was associated with RNFL thinning in POAG.