PURPOSE
To evaluate the association between rates of choroidal microvasculature dropout (MvD) change and rates of circumpapillary retinal nerve fiber layer (cpRNFL) loss in primary open-angle glaucoma (POAG) eyes.
DESIGN
Cohort study from clinical trial data.
METHODS
A total of 91 eyes of 68 POAG patients with and without localized MvD at baseline with at least 4 visits and 2 years of follow-up with optical coherence tomography angiography (OCT-A) and OCT scans were included. Area and angular circumference of MvD were evaluated on OCT-A en face and B-scan choroidal vessel density images during the follow-up period. Joint longitudinal mixed effects models were used to estimate the rates of change in MvD area or angular circumference and RNFL thickness. Univariable and multivariable regressions were completed to identify the factors contributing to cpRNFL thinning.
RESULTS
MvD was identified in 53 eyes (58.2%) at baseline. Seventeen eyes (18.6%) that did not show MvD at baseline developed it over the follow-up period. Over a mean follow-up of 4.0 years, the mean rates of change in MvD area and angular circumference (95% CI) were 0.05 (0.04, 0.06) mm 2 per year and 13.2° (10.7°, 15.8°) per year, respectively. In multivariable models, the rate of cpRNFL thinning was significantly associated with the rates of change in MvD area and angular circumference ( P = .008 and P = .009, respectively).
CONCLUSIONS
Rates of MvD area and angular circumference change over time were associated with concurrent rates of cpRNFL loss in POAG eyes.
G laucoma is a progressive optic neuropathy characterized by structural changes in the optic nerve head (ONH) and circumpapillary retinal nerve fiber layer (cpRNFL) and accompanying damage in the visual field (VF). , Although the pathophysiological process in glaucoma is still unclear, the vascular system and, in particular, the retinal and choroidal microvasculature, seem to have a contributory role. ,
A localized parapapillary perfusion defect, known as choroidal microvasculature dropout (MvD), has been well recognized in the choroidal layer of glaucoma patients on optical coherence tomography angiography (OCT-A) scans. MvD was found to correspond to perfusion defects on indocyanine green angiography, indicating that MvD represents a true circulatory defect in the peripapillary choroid related to a compromised perfusion in the deep tissues of the optic nerve head (ONH). , When detected at baseline, MvD was associated with progressive VF loss and/or cpRNFL thinning. Moreover, its area topographically corresponded with the location of cpRNFL defects in the majority of the eyes. , Recently, an enlargement of peripapillary MvD was observed in glaucomatous eyes, and these changes in size appeared to be greater in eyes with accelerated glaucoma progression.
Disc hemorrhage (DH) and decreased ocular perfusion pressure have been strongly associated with the presence of MvD. Other studies have reported a topographic association between MvD and LC defects. , , This suggests that a microvasculatory impairment involving LC and peripapillary choroid, both supplied by the short posterior ciliary artery, may explain the structural changes in the LC and lead to MvD. Localized perfusion defects in the peripapillary choroid have been more frequently seen in eyes with initial parafoveal VF defects within a 10° radius compared to eyes with initial nasal VF damage. In addition, systemic risk factors, such as diastolic blood pressure, cold extremities, and migraine, known to be associated with parafoveal scotomas, have been associated with MvD. , These findings suggest that the presence of MvD may be part of a specific pattern of glaucomatous central VF loss with an underlying pathogenetic mechanism of glaucomatous optic neuropathy. The loss of sensitivity in the central area of the VF carries a greater impact on the ability to perform activities of daily living in patients diagnosed with glaucoma. Therefore, its progressive decline is closely related to faster decline in the quality of life of glaucoma patients.
The current study aims to evaluate whether changes in MvD over time are related to progressive cpRNFL thinning. This may enhance our understanding of the role of MvD in glaucoma monitoring and provide a rationale for using MvD changes as a potential marker of glaucoma progression.
METHODS
This longitudinal study included POAG patients enrolled in the Diagnostic Innovations in Glaucoma Study (DIGS) , who underwent OCT-A (Angiovue; Optovue Inc) and spectral-domain OCT (Spectralis, Heidelberg Engineering, Inc) imaging. Participants were assessed longitudinally according to a standard protocol consisting of regular follow-up visits with clinical examination, imaging, and functional tests. All participants from the DIGS study who met the inclusion criteria described below were included. Informed consent was obtained from all study participants. The University of California, San Diego Human Subject Committee approved all protocols, and the methods described adhered to tenets of the Declaration of Helsinki.
This study included eyes with a minimum of 4 qualified OCT-A and OCT scans of ONH and a minimum of 2 years follow-up. Eyes were classified as having glaucoma if they had repeatable (at least 2 consecutive) abnormal VF test results with evidence of glaucomatous optic neuropathy, defined as excavation, presence of focal thinning, notching of neuroretinal rim, or localized or diffuse atrophy of the cpRNFL on the basis of masked grading of optic disc photographs by 2 graders. An abnormal VF test result was defined as a pattern SD ( P < .05) or a Glaucoma Hemifield Test result outside normal limits.
Inclusion criteria also included the following: (1) age greater than 18 years, (2) open angles on gonioscopy, and (3) best-corrected visual acuity of 20/40 or better at study entry. Exclusion criteria were as follows: (1) history of trauma or intraocular surgery (except for uncomplicated cataract surgery or glaucoma surgery), (2) coexisting retinal disease, (3) uveitis, (4) non-glaucomatous optic neuropathy, and (5) axial length of 27 mm or more. Participants with a diagnosis of systemic diseases such as Parkinson disease, Alzheimer disease, dementia, or a history of stroke were excluded. Those with poor-quality OCTA and OCT were also excluded.
OCT MEASUREMENTS
The Spectralis SDOCT (Spectralis HRA+OCT, software version 5.4.7.0; Heidelberg Engineering, Inc) was used for cpRNFL thickness measurements. The details of the SDOCT imaging have been previously reported. The cpRNFL thickness was measured using a high-resolution cpRNFL circle scan consisting of 1536 A-scan points from a 3.45-mm circle centered on the optic disc. Follow-up scans were performed using a built-in automated realignment procedure (referred to as the follow-up examination in the system documentation). The accuracy of the segmentation of the cpRNFL was reviewed, and segmentation errors were fixed manually by Imaging Data Evaluation and Analysis Reading (IDEA) Center personnel.
CHOROIDAL MICROVASCULATURE DROPOUT DETECTION
ONH 4.5 × 4.5 mm 2 (304 B-scans × 304 A-scans per B-scan) centered on the ONH were acquired with the AngioVue OCT-A system (software version 2018.1.1.63). The retinal layers of each scan were segmented automatically by the AngioVue software, and the en face choroidal vessel density map was acquired. Only good-quality images were included. OCT-A and SD-OCT image quality review was completed according to the IDEA Center standard protocol on all scans processed with standard AngioVue software. Poor-quality images were excluded; these were defined as images with the following: (1) low scan quality with quality index (QI) of less than 4; (2) poor clarity; (3) residual motion artifacts visible as irregular vessel pattern or disc boundary on the en face angiogram; (4) image cropping or local weak signal resulting from vitreous opacity; or (5) segmentation errors that could not be corrected. Dropout was required to be present in at least 4 consecutive horizontal B-scans and also to be greater than 200 µm in diameter in at least 1 scan and to be in contact with the OCT disc boundary. The optic disc boundary was automatically detected by the Optovue software. In case of errors in disc demarcation, 1 trained observer masked to the clinical information of the subjects corrected the disc boundary manually by searching for the position of the Bruch membrane opening, as previously described. Two observers who were masked to the clinical characteristics of the participants independently determined the presence or absence of MvD for each patient.
ASSESSMENT OF MvD AREA, CIRCUMFERENTIAL ANGLE, AND LOCATION
The optic disc and the peripapillary atrophy (PPA) margins were detected by simultaneously viewing the stereoscopic optic disc photographs and the scanning laser ophthalmoscopic (SLO) images that were obtained along with the OCT-A images. The MvD area was manually demarcated on en face choroidal vessel density maps using the line tool provided by ImageJ software, version 1.53 (National Institutes of Health; http://imagej.nih.gov/ij/download.html). The Littmann formula was used to correct the ocular magnification in OCT-A. , Details of the formula are provided elsewhere. The Avanti SD OCT has a default axial length of 23.95 mm and an anterior corneal curvature radius of 7.77 mm.
The MvD angular circumference was measured as previously described. In brief, the 2 points at which the extreme borders of MvD area met the ONH border were identified and defined as angular circumferential margins. The angular circumference was then determined by drawing 2 lines connecting the ONH center to the angular circumference margins of the MvD.
Both area and angular circumference of the MvD were assessed by 2 trained graders (E.M. and N.E.N.) who were masked to the clinical data of the patients, including cpRNFL data. Any uncertainty was resolved by a glaucoma specialist, designated as the third grader (S.M.). The MvD area that contained large retinal vessels was included as part of the MvD area if the MvD extended beyond the vessels. In cases in which the retinal vessels were located at the border of the MvD, the area covered by the vessels was excluded from the MvD area. Reflectance or shadowing of the large vessels on the horizontal and en face images was excluded from the quantitative analysis by the 2 independent graders. In an eye showing more than 1 MvD, the area and the angular extent of each MvD were calculated separately and then added together to determine the total area and the total angular extension of MvD for the eye.
The sectoral location of the dropout was determined based on the 8 separate sectors corresponding to those on the cpRNFL vessel density map of the OCTA. For each MvD, a line was drawn to equally bisect the angular circumferential margins of the MvD from the ONH center, as previously reported, to define the location of the MvD. Disagreements between the 2 observers in assessing the presence and location of the MvD were adjudicated by the third experienced grader (S.M.). The location of MvDs was categorized as hemispheric (superior, inferior, or both) corresponding to those on OCT-A and OCT parameters.
STATISTICAL ANALYSIS
Patient and eye characteristics data were presented as mean (95% CI) for continuous variables and as count (%) for categorical variables. Interobserver agreement in detecting the presence of MvD was assessed by using k statistics (ie, k value), and an interclass correlation coefficient (ICC) was used for the MvD area and angular circumference measurements. Categorical variables were compared using the χ 2 test.
Mixed-effects modeling was used to compare ocular parameters among groups. A joint longitudinal linear mixed effects model was used to estimate the rates of change for both cpRNFL and MvD, with random effects applied at the eye level. Details regarding the use of these models for evaluation of rates of change in glaucoma and to model longitudinal processes have been published. , Briefly, linear mixed models estimate the average rate of change in an outcome variable using a linear function of time, and subject- and eye-specific deviations from this average rate are introduced by random slopes, while taking into account for the potential influence of inter-eye correlation. In joint longitudinal modeling, both cpRNFL and MvD data are modeled simultaneously. This allows a better determination of the true underlying relationship between the 2 outcomes by considering measurement error. Specifically, we evaluated the association between cpRNFL rate of change with the MvD area and the MvD angular circumference rate of change.
The effect of rates of MvD area or angular circumference changes as well as potential predictors, such as age, mean intraocular pressure (IOP) during follow-up, and any other variable in which the P value was less than .1 on univariable analysis of the RNFL thinning during the follow-up was introduced in the multivariable model. Statistical analyses were performed using Stata version 16.0 (StataCorp). P values of less than .05 were considered statistically significant for all analyses.
RESULTS
A total of 101 eyes of 75 POAG patients were initially included. Among them, 10 eyes of 7 patients were excluded because of the presence of artifacts or shadowing from large vessels, resulting in 91 eyes of 68 patients that were finally included in the analysis.
The demographic characteristics are provided in Table 1 . Mean (95% CI) age was 69.0 (66.6, 71.4) years, and mean (95% CI) baseline VF MD was −4.7 (−5.8, −3.6) dB. The mean (95% CI) numbers of OCT and OCTA visits were 6.2 (5.8, 6.6), and 5.7 (5.4, 6.0), over a mean follow-up of 4.0 (3.9, 4.2) years. Interobserver agreement in detecting the presence of MvD (95% CI) was excellent (k = 0.988 (0.986, 0.990). The ICC for interobserver reproducibility in measuring the area and the angular circumference of MvD (95% CI) were 0.977 (0.973, 0.980) and 0.977 (0.973, 0.980), respectively.
| Variable | Value |
|---|---|
| Total no. of patients (eyes) | 68 (91) |
| Baseline age, y | 69.0 (66.6, 71.4) |
| Sex, female/male | 34/34 |
| Race, African American/non−African American | 19/49 |
| Self-reported hypertension, n (%) | 46 (66.7) |
| Self-reported diabetes, n (%) | 11 (15.9) |
| Axial length, mm | 24.1 (23.9, 24.4) |
| CCT, µm | 532.7 (523.3, 542.1) |
| Baseline IOP, mm Hg | 14.1 (13.2, 14.9) |
| Disease Severity by Baseline 24-2 VF MD | |
| Early glaucoma (MD ≥ −6 dB), no. of eyes (%) | 67 (73.6) |
| Moderate and advanced glaucoma (MD < −6 dB), no. of eyes (%) | 24 (26.4) |
| Baseline VF MD, dB | -4.7 (-5.8, -3.6) |
| cpRNFL at baseline, µm | 72.5 (69.3, 75.7) |
| cpRNFL at last visit, µm | 70.0 (66.9, 73.2) |
| Corrected MvD area at baseline, mm 2 | 0.15 (0.09, 0.22) |
| Corrected MvD area at last visit, mm 2 | 0.32 (0.24, 0.40) |
| MvD angle at baseline, degrees | 41.3 (29.3, 53.4) |
| MvD angle at the last visit, degrees | 85.4 (69.4, 101.4) |
| OCT follow-up visits, n | 6.2 (5.8, 6.6) |
| OCTA follow-up visits, n | 5.7 (5.5, 6.0) |
| Follow-up, y | 4.0 (3.9, 4.2) |
Peripapillary MvD was detected in 53 eyes (58.2%) at baseline, whereas 17 eyes (18.6%) that did not show MvD at baseline developed it over the follow-up period. Among the 53 eyes with MvD at baseline, in 21 eyes (39.6%) MyD was located in the inferior hemisphere, in 4 eyes (7.6%) in the superior hemisphere, and in 28 eyes (52.8%) in both hemispheres.
The mean (95% CI) MvD areas were 0.15 (0.09, 0.22) mm 2 at baseline and 0.32 mm 2 (0.24, 0.41) at final follow-up ( P < .001). The mean (95% CI) MvD angular circumference was 41.3° (29.3°, 53.4°) at the baseline, whereas it was 85.4° (69.4°, 101.4°) at the final follow-up ( P < .001). The mean (95% CI) rates of change in MvD area and angular circumference were 0.05 (0.04, 0.06) mm 2 per year and 13.22° (10.67°, 15.76°) per year, respectively. The rates of change in the MvD area in eyes with mean IOP during follow-up greater than 16 mm Hg were significantly lower compared to those with a mean IOP during follow-up of less than 16 mm Hg (0.02 [−0.03, 0.00] mm 2 /y lower, P = .049).
The Figure 1 shows a representative case of the relationship between MvD area enlargement and cpRNFL thinning in a glaucomatous patient.
Table 2 shows the effect of the corrected MvD area change on cpRNFL thinning in the univariable and multivariable mixed model. On univariable analysis, the rate of change in MvD area was significantly associated with cpRNFL thinning (−6.11 [−12.60, 0.38] µm/y per 1-mm 2 /y increase; P = .064. After adjustment for confounding factors (ie, age, self-reported hypertension, and mean IOP during follow-up), the rate of change in the MvD area and baseline VF MD (coefficient [95% CI) were significantly associated with cpRNFL progression (−8.80 [−15.24, −2.36] µm/y per 1-mm 2 /y increase, P = .008; and 0.06 [0.03, 0.09] dB/y per 1-dB worse, P < .001).
