Introduction

Microvascular changes of the optic nerve head are considered a potential factor in the pathogenesis of open-angle glaucoma (OAG) [1, 2]. This proposition has been supported by studies which showed that in eyes with glaucomatous damage, peripapillary choroidal microvasculature was more impaired than contralateral healthy eyes [3]. In this regard, peripapillary choroidal microvasculature dropout (MvD) has been of particular interest. This regional vasculature loss as shown on optical coherence tomography angiography (OCTA) exactly mirrors perfusion defect on indocyanine green angiography [4], and many investigators have previously argued that it may be a sign of impaired juxtapapillary choroidal blood flow and compromised perfusion in the prelaminar tissue [5]. This view is further strengthened by literature that show a higher prevalence of concurrent systemic risk factors, such as low diastolic blood pressure (DBP), cold extremities, and migraine, in eyes with MvD compared to eyes without the dropout [6, 7].

Its significance in the development and progression of glaucoma has been actively investigated. Studies that prove its association with retinal nerve fibre layer (RNFL) thinning and visual field (VF) defect abound [8,9,10]. However, while numerous studies reported that MvD is most frequently present in areas spanning from the inferior to temporal sectors [3, 11], few studies have analysed its locational significance. Of note, MvD’s have been observed in non-glaucomatous optic neuropathy, such as compressive optic neuropathy (CON) and non-arteritic ischemic optic neuropathy (NAION), chiefly clustered in the temporal sector [12, 13]. If the location of MvD in these diseases is any indication, the tendency of a MvD to be present in certain sectors over others suggests that the location of MvD in glaucomatous eyes may be related to specific pathoetiology. We hypothesized that different locations of MvD are associated with different patterns of primary neuronal damage in glaucoma. The purpose of this study was to compare and identify clinical characteristics associated with glaucoma progression in eyes with MvD in different locations.

Materials and methods

Patient population

The study protocol was approved by the Institutional Review Board of Severance Hospital at Yonsei University (IRB approval number 4-2022-0985) and the study was conducted in adherence to the tenets of the Declaration of Helsinki. Informed consent was waivered due to the retrospective nature of the study. A total of 181 consecutive eyes with diagnosis of OAG were enrolled following a retrospective review of patients who were referred to the Department of Ophthalmology, Severance Hospital between March 2017 and December 2020. Patients who were 40 years or older at the time of OAG diagnosis were considered eligible for inclusion. OAG was diagnosed when glaucomatous optic disc changes (such as localized or diffused neuroretinal rim thinning, a difference of cup-to-disc ratio greater than 0.2 between 2 eyes) were noted on stereo-disc photographs, RNFL defect was identified on either red-free fundus photographs or Cirrus OCT (Carl Zeiss Meditec, Inc., Dublin, California) and open angle was confirmed with gonioscopy. Two glaucoma specialists (JSL and WC) re-evaluated the diagnosis of OAG and any discrepancy was resolved by a third adjudicator (HWB). Standard automated perimetry (SAP; 24-2 SITA standard, Humphrey Field Analyzer II; Carl Zeiss Meditec, Inc., Dublin, California) was also performed. Patients with advanced stages of glaucoma as per Hodapp-Parrish-Anderson criteria (VF defects with mean deviation (MD) <-12 dB and either ≥50 points on the pattern standard deviation (PSD) plot depressed below the 5% level and ≥25 but <50% of points depressed below the 1% level, or points within the central 5° with a sensitivity <15 dB in both hemifields on reliable VF tests) were excluded from the analysis in order to avoid a “floor effect” in measurements of RNFL and ganglion cell-inner plexiform layer (GCIPL) thicknesses [14, 15]. A VF test was considered reliable when fixation losses were less than 33% and false positive rates were below 33%. A VF defect was considered to be associated with glaucoma if the glaucoma hemifield test results were abnormal or if 3 contiguous VF locations on PSD were below 5% significance levels and if the defects were reproduced in the same location on 2 consecutive tests. A minimum of 5 ‘reliable’ VF tests per eye was required for evaluation of VF defect progression. On average, 7.5 ± 2.3 VF tests were performed for each eye over 7.3 ± 4.0 years of follow-up (7.6 ± 2.5 tests for inferior MvD eyes vs. 7.1 ± 2.2 tests for temporal MvD eyes; P = 0.093). A total of 6 eyes (3 with temporal MvD and 3 with inferior MvD) were excluded from analysis due to insufficient numbers of reliable VF tests.

Subjects satisfying any of the following criteria were excluded from the study: 1) eyes without β-peripapillary atrophy (PPA) on fundus photographs or OCT; 2) eyes with parapapillary γ-zone with a maximum horizontal width exceeding 200 μm on infrared imaging (because a wide parapapillary γ-zone may be associated with peripapillary microvascular impairment from scleral stretching) [16]; 3) refractive error less than −8.00 D or greater than +3.00 D; 4) significant media opacity such as cataract; 5) clinical evidence of intracranial lesion, neurologic disorder, rheumatologic disease or systemic vasculitis; 6) systemic medication known to induce optic neuropathy (e.g. ethambutol, digoxin and vigabatrin); 7) history of optic neuritis or NAION; 8) history of intraocular surgery other than uncomplicated cataract surgery; 9) history of ocular disease that may increase IOP and cause secondary glaucoma; 10) ocular manifestation of systemic disease known to affect optic disc such as diabetic retinopathy and retinal vessel occlusion; 11) presence of pseudoexfoliation or pigment dispersion; 12) history of ocular trauma; 13) failure to produce 2 consecutive reliable VF tests at baseline; and 14) less than 5 high signal strength (high signal strength was defined as ≥ 7/10) cirrus OCT tests. When both eyes were eligible, the right eye was chosen.

Ophthalmologic examinations

All subjects underwent complete ophthalmologic examinations at baseline, including medical history, visual acuity, intraocular pressure (IOP; measured with Goldmann applanation tonometer (GAT), Haag-Streit model BQ-900; Haag-Streit, Inc., Bern, Switzerland), refraction, slit-lamp examinations, gonioscopy, central corneal thickness (CCT), axial length, dilated fundus examinations, colour disc stereophotography, red free fundus photography, SAP, cirrus OCT and cirrus OCT angiography (AngioPlex; Carl Zeiss Meditec, Inc., Dublin, CA). Patients were followed every 6 months, undergoing visual acuity, IOP measurements, slit-lamp examinations, colour disc stereophotography, red free fundus photography, SAP and cirrus OCT each visit. Zeiss Forum software (Carl-Zeiss Meditec, Dublin, California, USA) was used to calculate progression rates for MD (dB/year) and visual field index (VFI; %/year). The average value of 2 repeated IOP measurements taken during the first 2 consecutive visits were considered the baseline IOP. The mean IOP is the average value of IOP’s collected every 6 months. IOP fluctuations were defined as the standard deviation (SD) of IOP’s collected every 6 months.

Optical coherence tomography and angiography

The thicknesses of RNFL and GCIPL were obtained using the 200 × 200 optic disc cube and 512 × 128 macular cube protocols of cirrus OCT version 6.0 software. The rates at which RNFL and GCIPL thicknesses changed during follow-up were determined using the trend-based algorithm of Guided Progression Analysis of Cirrus OCT. The trend-based algorithm is designed to align superpixels (4 × 4 pixels), measure the thickness of each superpixel and perform linear regression analyses in order to calculate changes in the thickness over the follow-up period. The 2 initial examinations were set as the baseline. All OCT scans were evaluated for image quality by a single investigator (JSL) and those with any artifacts were excluded from analysis.

The retinal vessel density (rVD) of the peripapillary and macular areas were obtained using a commercially available OCT angiography device. The angiography algorithm is described in detail elsewhere [17]. The AngioPlex software employed in this study calculated rVD by adding the total length of perfused vasculature per unit area of a 6mm-diameter circle placed on en face images of peripapillary and macular inner-retinal layer, which was automatically segmented by the OCT software. Any segmentation error was manually corrected and all scans were examined for quality by a single investigator (JSL). Those scans with any motion artifacts or low signal strength (<7/10) were excluded.

Determination of peripapillary choroidal MvD

An MvD was defined as a focal and complete loss of choriocapillaries or microvasculature within β-PPA on en face images of the choroid. The choroidal layer was manually generated by creating an en-face slab that extends from the retinal pigment epithelium to 390 μm below, which is a distance considered sufficient to include both the choroid and the inner scleral layer [12]. A microvasculature loss was defined as the absence of choriocapillaries adjoining the disc margin for a circumferential width greater than a half clock-hour. The location of MvD was determined based on the 4 circumferential quadrants of sectoral RNFL thickness measurements of the cirrus OCT. If an MvD spanned over more than 1 neighbouring sector, the sector that included larger portions of the MvD was assigned. The presence and the location of MvD were determined by 2 independent observers (JSL and SP), who were blinded to the clinical information of subjects. Any discrepancy was resolved by a third adjudicator (HWB). Based on the location of the peripapillary choroidal MvD, patients were assigned to either inferior MvD or temporal MvD groups.

Statistical analysis

Interobserver agreement on the presence and location of MvD was analysed using the κ statistics. The distribution of data was determined with Wilk-Shapiro test. Continuous data are presented as mean ± SD, and categorical data as number (percentage). Independent samples t-test and Mann-Whitney U test were used for comparisons of normally distributed and non-normally distributed continuous data, respectively. Chi-square test was used for comparisons of categorical variables. The mean rates of change in RNFL and GCIPL thicknesses between inferior and temporal MvD groups were compared using a linear mixed model that adjusted for age, average IOP, baseline VF MD, baseline GCIPL and RNFL thicknesses. Univariate linear regression analyses were performed separately for each variable. To build a multivariate model, a stepwise selection method was adopted with the entry P-value of <0.1 and a stay P-value of <0.05. All statistical analyses were conducted using SPSS version 23 (SPSS Inc., Chicago, Illinois). A p-value < 0.05 was considered statistically significant.

Results

Baseline characteristics

The clinical characteristics of OAG eyes with inferior and temporal MvD are illustrated in Table 1. MvD was identified mainly in the temporal quadrant in 110 eyes (temporal MvD; 60.5 ± 12.6 years, 39.1 % males), and in the inferior quadrant in 71 eyes (inferior MvD; 60.3 ± 11.1 years, 39.4 % males). The κ coefficient for interobserver agreement on the presence of MvD was 0.918. The intraclass correlation coefficient (ICC) for interobserver reproducibility in determining the location of the MvD was 0.899. The follow-up duration was comparable between the 2 groups (P = 0.871). The 2 groups were similar in age, gender, axial length and CCT. The 2 groups did not show significant differences in IOP at baseline or during follow-up. Baseline VF parameters were also comparable. Inferior RNFL of the inferior MvD group (73.8 ± 19.0 µm) was thinner than that of the temporal MvD group (80.3 ± 21.9 µm; P = 0.041) at baseline. Superior (P = 0.010), superonasal (P = 0.001) and superotemporal (P = 0.014) GCIPL was thicker for the inferior MvD group than the temporal MvD group at baseline. No significant differences were found in underlying systemic medical conditions such as hypertension and diabetes mellitus.

Table 1 Baseline characteristics of patient population depending on the location of MvD.
Full size table

Comparison of rates of RNFL and GCIPL thinning

The rates of global and regional RNFL and GCIPL thinning were comparable between eyes with inferior and temporal MvD (Supplemental Table 1). The 2 groups were further divided into small and large long-term IOP fluctuations using the median value of 1.7 mmHg (Table 2). IOP fluctuations were defined as the SD of IOP collected during office visits every 6 months. Significantly faster mean rates of RNFL and GCIPL thinning were found in the inferior hemifield in eyes with inferior MvD compared to temporal MvD eyes when the long-term IOP fluctuations were above the median value after adjusting for confounding factors, including age, baseline VF MD, mean IOP, baseline RNFL thickness and baseline GCIPL thickness (mean difference (95% CI); −0.833 (−1.298 to −0.367) µm/year, P = 0.022; −0.741 (−0.952 to −0.530) µm/year, P = 0.020, respectively). However, the difference was not noted in comparisons of VF progression rates between the 2 groups (Supplemental Table 2). The rates of RNFL and GCIPL thinning were comparable between the 2 groups when IOP fluctuations were small. The mean rates of RNFL and GCIPL thinning were also similar when the groups were divided using baseline IOP or average IOP (Supplemental Table 3).

Table 2 Comparison of rates of structural progression between inferior and temporal MvD depending on long-term IOP fluctuation.
Full size table

Factors contributing to RNFL thinning in MvD eyes

Factors associated with RNFL thinning in superior and inferior hemifields in eyes with MvD were identified (Table 3). According to the multivariable analysis, large IOP fluctuations were significantly associated with faster rates of RNFL thinning in the inferior hemifield in inferior MvD eyes (β −0.463, 95% CI −1.111 - −0.264; P = 0.002), but not in temporal MvD eyes (P = 0.549). Elevated baseline IOP was associated instead with RNFL thinning in the inferior hemifield of eyes with temporal MvD (β 0.279, 95% CI 0.014–0.228; P = 0.027).

Table 3 Linear regression analysis to identify factors associated with RNFL thinning in different MvD groups.
Full size table

Discussion

The present study demonstrated that the rates of RNFL and GCIPL thinning in the inferior and superior hemifields were comparable between eyes with temporal and inferior MvD. However, the thinning rates in the inferior RNFL and GCIPL were significantly faster in eyes with inferior MvD in comparison to eyes with temporal MvD in the presence of large long-term IOP fluctuations. Long-term IOP fluctuations were independently associated with inferior RNFL thinning in eyes with MvD in the inferior quadrant. The findings of this study suggest that MvD in different locations may be associated with different patterns of glaucomatous progression.

Numerous studies in the past have linked MvD to glaucoma severity and progression [18, 19]. For instance, MvD has shown significant associations with thinner RNFL and worse VF MD in one study [6]. Another study found that the dropout was more frequently present in eyes with central VF defect [20]. Its presence as well as its area and angular circumference were found to be independently related with faster RNFL thinning [19, 21]. Increases in size of the dropout over time were also significant markers of faster rates of RNFL thinning as well as VF loss [22, 23]. Previous studies have not come to terms on whether MvD is a primary pathogenic phenomenon or a secondary consequence of glaucomatous damage [5, 24], but these studies do generally agree that MvD helps identify high-risk glaucoma. We believe the present study, by comparing the rates of RNFL and GCIPL thinning among eyes with MvD in different locations, adds to the current literature that MvD location may also be relevant to glaucoma pathogenesis.

The results suggest that the location of MvD may reveal different primary pathogenic mechanisms of glaucomatous damage. Although the site of MvD did not affect the thinning rates in general, eyes with inferior MvD showed faster RNFL and GCIPL thinning in the same hemifield than eyes with temporal MvD when large long-term IOP fluctuations were present. With respect to OAG in general, there exist 2 major theories on the pathogenesis: the mechanical and vascular theories. The mechanical theory posits that elevated IOP puts mechanical stress on retinal ganglion cells to cause irreversible damage [25]. This theory may be especially applicable to damages in the superior and inferior hemispheres because the superior and inferior portions of the lamina cribrosa tend to have larger pores and less connective tissue support compared to the nasal and temporal portions [26], and mechanical insult from elevated IOP is likely exaggerated in these regions. According to the results of our study, eyes with MvD in this susceptible region showed faster RNFL and GCIPL thinning in the same hemisphere when large IOP fluctuations were also present. The relationship between the localized choroidal perfusion defect (MvD) and IOP fluctuations is not entirely clear. However, previous studies have shown that MvDs are more susceptible to large IOP fluctuations than elevated mean IOP. For instance, Jo et al. described that glaucoma eyes with MvD tended to experience greater long-term IOP fluctuations [10]. Micheletti et al. have reported faster MvD enlargement in both area and angular circumference when in the presence of large IOP fluctuation, and not in that of  high average IOP [23]. In the same study, eyes that developed MvD during follow-up tended to show large IOP fluctuations compared to eyes that did not develop any MvD during the same period. Other studies have also found that peak IOP tended to be higher in eyes with MvD [7]. Based on these studies, we can extrapolate that repeated loading and unloading of stresses on the fibroelastic tissues of the lamina cribrosa from IOP fluctuations, when combined with impaired choroidal perfusion, resulted in more rapid structural progression in our patient population. It is also possible that the damaging effect of IOP fluctuations may be more apparent in a study population with low mean IOP such as our own, whose average IOP was in the low teens [27]. However, further investigations are necessary to establish the relationship between choroidal perfusion, IOP fluctuations and structural progression in glaucoma.

The vascular theory, on the other hand, states that some of glaucoma are associated with alterations in ocular blood flow [2]. In other words, retinal hypoperfusion results in reduced oxygen supply and predisposes metabolically active retinal ganglion cells to injury. A number of optic neuropathy in part share this pathogenic mechanism with glaucoma, namely CON and NAION [12, 13]. They involve acute hypoperfusion of the short posterior ciliary artery, resulting in severe ischemia in the distal branches [28, 29]. MvD has also been reported in these diseases, and interestingly enough, it was most frequently observed in the temporal parapapillary sector in association with thinner RNFL and reduced retinal vessel density. We speculate that the temporal MvD may be present in glaucoma in association with this mechanism. However, while the temporal MvD eyes were relatively less affected by IOP fluctuations in the present study, they did not show characteristic features of the subtypes of glaucoma thought to be affected by impaired perfusion to affirm our speculation. For instance, past investigations have related these subtypes of glaucoma with predominant macular GCIPL loss and more central VF defect [30, 31]. Furthermore, when Yum et al. compared clinical characteristics between temporal and inferotemporal RNFL defect among normal-tension glaucoma (NTG), eyes with temporal RNFL defect tended to show lower best-corrected visual acuity and more prevalent central scotoma [32]. The temporal MvD eyes in the present study did not show significantly lower VFI values and GCIPL thicknesses than those of the inferior MvD eyes. However, we do suspect that temporal MvD was more prevalent than inferior MvD in our study population in part due to the equally high prevalence of NTG, which is generally believed to result from regional vascular insufficiency. Nonetheless, larger samples with greater variations in baseline IOP are necessary to assess our speculations regarding the temporal MvD.

According to our analysis, the difference in structural progression rates between inferior MvD and temporal MvD eyes did not result in different functional deterioration rates (Supplemental Table 2). With regards to these results, we propose following explanations. First, damaged retinal ganglion cells might have been compensated by a complex network of neural cells, resulting in a VF test that does not directly reflect the amount of nerve damage [33]. Furthermore, it is possible that current OCT provides incomplete information on cells that are intact in structure, but damaged in function, and vice versa [34]. Second, a more sensitive test of vision performance might have been necessary to identify any correlation between VF defect and different structural damage rates between the 2 groups. Inter-individual variabilities in structure-function relationships, and discordance between the region of retina tested by a VF stimulus and measured by OCT may further contribute to the lack of difference in VF in our analysis [15, 35, 36]. Third, a significant difference in VF damage might have been noted if the data were collected for a longer period. Previously, a time lag was reported to exist between initial nerve fibre damage and resultant VF defect [37]. Further studies with a longer follow-up are necessary to prove our speculations.

There are several limitations to the study. First of all, the study is limited by its retrospective design. Second, the study includes a small number of subjects, and the small sample size may have affected the results. Third, the study results may not be generalized as the study population consisted of Koreans, among whom NTG is more prevalent. Fourth, projection artifacts of superficial vessels may have created false images of vascularity in the deep retinal layers, resulting in false negative detection of MvD. Lastly, the study results may not be generalizable to glaucomatous eyes without β-PPA, and further investigations are necessary to identify whether certain regions of RNFL or GCIPL are more prone to damage in the presence of large IOP fluctuations over others.

In conclusion, the rates of RNFL and GCIPL thinning are comparable between glaucomatous eyes with temporal and inferior MvD. However, structural progression is more rapid in eyes with MvD in the inferior hemifield when accompanied by long-term IOP fluctuations, but not in eyes with MvD in the temporal sector. These findings suggest that progression of glaucoma in association with MvD in different locations are affected by different risk factors. MvD may be useful for detection of patients at a high risk of rapid progression who require close observation and strict IOP control.

Summary

What was known before

  • Choroidal microvasculature dropout (MvD) as shown on optical coherence tomography angiography (OCTA) exactly mirrors perfusion defect on indocyanine green angiography.

  • MvD may be a sign of impaired juxtapapillary choroidal blood flow and compromised perfusion in the prelaminar tissue.

What this study adds

  • Eyes with inferior MvD showed faster RNFL and GCIPL thinning in the same hemifield than eyes with temporal MvD when large long-term IOP fluctuations were present.

  • The location of MvD in glaucomatous eyes may be related to specific pathoetiology.