Association of optic nerve head prelaminar schisis with glaucoma

Abstract

Purpose:

To compare the frequency of observing optic nerve head (ONH) prelaminar schisis by optical coherence tomography (OCT) in glaucoma and glaucoma suspect (GL/S) eyes versus healthy control (HC) eyes and to assess its association other markers of glaucoma severity.

Methods:

This cross-sectional study included 298 eyes of 150 GL/S patients and 88 eyes of 44 HC. OCT scans were obtained by SPECTRALIS (Heidelberg Engineering GmbH), including 24 radial B-scans, each comprised of 768 A-lines spanning 15°, centered on the ONH. Two reviewers masked to all other clinical, demographic and ocular information, independently graded the OCT scans for the presence of ONH prelaminar schisis on a 4-point scale: 0 (none) to 3 (severe). The probability of ONH schisis was compared between groups and against demographic and ocular factors, including structural and functional measures of glaucoma severity.

Results:

The frequency and severity of ONH prelaminar schisis were greater in GL/S than in HC (p=0.009). Among the GL/S group, 165 eyes (55.4%) had no visible schisis (Grade-0), 71 (23.8%) had Grade-1, 46 (15.4%) had Grade-2 and 16 (5.4%) had Grade-3 schisis. Among HC eyes, 59 (67.0%) had Grade-0, 24 (27.3%) had Grade-1, 5 (5.7%) had Grade-2, none had Grade-3. ONH schisis was more common in eyes with thinner MRW and a deeper cup.

Conclusions:

ONH prelaminar schisis may be a sign of glaucomatous deformation and reflect ongoing pathophysiological damage. ONH prelaminar schisis can impact OCT image segmentation and diagnostic parameters, resulting in substantial overestimation of the true rim tissue thickness and underestimation of cup depth.

Table of Contents Statement:

Schisis within the prelaminar optic nerve head is detected by optical coherence tomography more commonly in glaucomatous eyes than in healthy eyes. The frequency of observing prelaminar schisis increases with worse glaucoma severity, particularly in eyes with thinner optic disc rim tissue and deeper optic cup. Its presence can impact image segmentation and diagnostic parameters, resulting in substantial overestimation of the true rim tissue thickness and underestimation of cup depth.

The optic nerve head (ONH) is considered to be the primary site of damage to retinal ganglion cell axons in glaucoma.^1–3 Indeed, the clinical appearance of the ONH in glaucoma is unique among optic neuropathies in that it generally manifests a greater degree of “cupping” and “excavation” for a given severity of vision loss and/or axon loss, which is thought to result from deformation and remodeling of the major load-bearing connective tissues within and surrounding the ONH (the lamina cribrosa and peripapillary sclera).^3–5 The distinct clinical appearance observed in glaucoma serves as an important diagnostic sign by which to differentiate it from other optic neuropathies.^{4, 6,7}

These ONH conformational changes exist along a spectrum throughout all stages of glaucoma and are often the underlying basis for considering some eyes (or persons) to be “glaucoma suspect” (e.g., glaucomatous appearance of the ONH without detectable visual field loss). Longitudinal change of the ONH toward an increasingly glaucomatous appearance is known to be a strong risk factor for subsequent development of glaucomatous visual field loss, that is, for conversion from suspect to manifest glaucoma.^{8, 9} Thus, even during early stages of glaucoma, the mechanical stresses thought to be driving deformation and remodeling of the ONH are also likely exerting mechanical and physiological stress and on adjacent tissues, perhaps compounded by the ONH conformational changes themselves. For example, we recently reviewed an array of findings from optical coherence tomography (OCT) studies, which demonstrated anatomical disruption within the retina of glaucomatous eyes, sometimes several millimeters from the ONH.¹⁰ One potential explanation we proposed is that the mechanical forces driving ONH deformation and remodeling (e.g., in glaucoma and myopia), along with the resulting conformational changes themselves, exert an impact on distant tissue via the retinal blood vessels, internal limiting membrane (ILM) and macroglia (Müller cells).¹⁰ In that review article, we also reported several cases in which the ONH prelaminar tissue had exhibited OCT signs of “schisis” or splitting in association with variants of retinal anatomical disruption in glaucomatous eyes, such as peripapillary retinoschisis and paravascular defects.¹⁰

While it will likely be informative to eventually determine whether there is an association between those (or other) retinal abnormalities and the ONH prelaminar schisis sign, it is more immediately important to determine if ONH prelaminar schisis occurs more frequently in glaucomatous eyes than in healthy eyes and whether its presence and/or apparent severity is associated with the degree of glaucomatous damage. Therefore, the purpose of this study was to compare the frequency of observing ONH prelaminar schisis by OCT in glaucoma and glaucoma suspect eyes versus healthy control eyes and to assess the association between ONH prelaminar schisis and other markers of glaucoma severity. Clinicians and researchers would use this information to better understand the spectrum of anatomical and pathological changes in glaucoma, and to recognize how prelaminar schisis may influence the OCT evaluation of glaucoma.

Methods

Participants

This study was conducted using data obtained from two groups of participants tested at the Devers Eye Institute in Portland, OR. The healthy control group consisted of 44 adults (88 healthy control eyes) who had been screened by complete ocular examination at the Devers Eye Institute and enrolled in a separate multicenter study designed to establish normative reference ranges for diagnostic OCT parameters used in commercial instrumentation.^{11, 12} This group thus represents a subset of the individuals comprising the Caucasian normative database described in detail in other reports.^{11, 12} The second group is comprised of participants enrolled in the ongoing Portland Progression Project (“P3”), a prospective longitudinal study whose aims include identification of risk factors for glaucoma progression and predictors of its rate.¹³ All aspects of this study adhered to the tenets of the Declaration of Helsinki, were approved and monitored by the Legacy Health Institutional Review Board and are compliant with the Health Insurance Portability and Accountability Act of 1996 (HIPAA). All participants provided written informed consent once all of the risks and potential benefits of participation were explained to them.

At enrollment into the P3 study, all participants had either a clinical diagnosis of early-stage primary open angle glaucoma (POAG), or their physician had determined that they had an optic disc appearance suspicious for glaucomatous optic neuropathy (large cup to disc ratio, cup-to-disc asymmetry ≥ 0.2 between eyes, rim notching, nerve fiber layer thinning or defect, history of disc hemorrhage) and at least one other risk factor for glaucoma (e.g. ocular hypertension with documented intraocular pressure ≥ 22mmHg on at least two occasions and/or first-degree family history of POAG). We also excluded participants from enrolling if visual acuity was worse than 20/40 in either eye and/or if there were any indicators of other ocular diseases or medications that could lead to changes in visual acuity or visual fields. Since their initial enrollment into the longitudinal P3 study, some eyes have progressed to develop worse glaucoma, yet many are still considered by their referring physicians to be glaucoma suspects. Thus, herein we refer to this group of eyes using the abbreviation “GL/S”. All participants had a complete ophthalmic examination of both eyes including measurement of visual acuity, slit lamp evaluation of the anterior segment, manifest refraction, gonioscopy, pachymetry, and evaluation of the ocular fundus, including optic disc and macula.

To be included in the Healthy Control group, all the following inclusion criteria were required: age between 18 and 90 years; clinically normal eye examination without clinically significant vitreo-retinal or choroidal disease or prior intraocular surgery except cataract or refractive surgery; intraocular pressure ≤ 21 mm Hg; best corrected visual acuity ≥ 20/40; refractive error within 6 D spherical error and 2 D astigmatic error; normal visual field with both the Glaucoma Hemifield Test (GHT) and Mean Deviation (MD) being within normal limits (see below).^{11, 12}

Visual Field Testing

Standard visual field testing was performed using a Humphrey Field Analyzer (HFA II, Carl-Zeiss Meditec, Dublin, CA, USA) with a size-III stimulus, SITA standard algorithm, and 24–2 test pattern. All visual field tests included in this study were considered reliable with fewer than 20% false positives and fewer than 33% fixation losses. We used the visual field GHT results along with the global indices MD and Pattern Standard Deviation (PSD) to evaluate associations between the severity of glaucomatous visual field loss and the presence or severity of ONH schisis (see below).

OCT Imaging

All OCT scans were obtained using a Spectralis instrument (Heidelberg Engineering, GmbH, Heidelberg, Germany) on the same day as visual field testing. The scan pattern was comprised of 24 radially oriented B-scans and three circumpapillary B-scans as selected from the “ONH-RC” preset available within the instrument’s Glaucoma Module Premium Edition. During acquisition for each eye in real time, the scan pattern is centered over the ONH and rotated into alignment with the fovea using the instrument’s proprietary Anatomic Positioning System. This system relies on an interactive routine whereby the operator identifies the position of the Bruch’s membrane opening (BMO) and the base of the foveal pit, each within a pair of live, orthogonal B-scans. Thus, the axis between the fovea and the BMO centroid (the FoBMO axis) defines the “temporal” position (at 0 degrees) of the pattern. The ONH radial B-scans were equally spaced 7.5 degrees apart; each contained 768 A-lines and spanned 15 degrees laterally (approximately 4.4 mm in an ‘average’ adult human eye). For this study, we used only the innermost of the three circular B-scans, which had a diameter of 12 degrees (approximately 3.5 mm) and contained 768 A-lines. The number of B-scan sweeps averaged to obtain the final recording in each case was predetermined by the Automatic Real Time (ART) Mean setting of 25 for ONH radial scans and 100 for the circumpapillary B-scan. All OCT scans had a quality score >20.

OCT image segmentations were inspected by the operator upon completion of scan acquisition and corrected manually if necessary.¹⁴ For the quantitative analyses applied in this study, we required accurate delineation of two image features within the ONH radial B-scans and two image features within each of the circumpapillary B-scans, as shown in Figure 1A–B and 1E, respectively. For radial scans, the inner limiting membrane (ILM, yellow contours in Figs. 1A–C) and Bruch’s membrane opening (BMO, red dots in Figs 1A–C) were used as described below.

Figure 1. Example of an individual GL/S eye from the P3 study group demonstrates methods used to quantify parameters from OCT scans.

A. OCT B-scan through the vertical meridian of the ONH (orthogonal to the FoBMO axis). The inset shows the B-scan location indicated by the bold green line overlaid onto the corresponding infrared reflectance image. Structures delineated in each of the 24 radial B-scans include the ILM (yellow contour) and BMO points (red dots). The green line segments connecting BMO points to the ILM represent the minimum distance from each BMO point to the ILM, i.e., the pair of MRW vectors determined for each B-scan. B. B-scan through the horizontal meridian of the ONH (along the FoBMO axis); note the MRW angle is shallow on the nasal (right) side. C. ILM and BMO delineations for all 24 radial B-scans shown from the perspective looking toward the inferior pole of the ONH. D. The parameter MRA was derived as the sum of the 48 radial trapezoidal sectors shown as a red, ribbon-like structure from the same perspective (looking toward the inferior pole of the ONH); note that the MRA Angle is shallow along the nasal rim. E. Corresponding circumpapillary B-scan showing delineations of ILM (yellow contour) and posterior RNFL boundary (red contour). F. Infrared reflectance image captured during the OCT scan has been laid onto the ILM surface to provide a clinical view of the surface topography in this eye; it is shown from the same perspective as the three-dimensional MRW and MRA representations in panels C and D, respectively; the view is toward the inferior pole of the ONH (temporal sector to the left, nasal to the right). Female, age=62 years; visual field MD=1.7 dB, PSD=1.2 dB, GHT=WNL; BMO Area=1.99 mm²; MRW=261 μm; MRW Angle=34 degrees; MRA=1.12 mm2; Mean Cup Depth=385 μm; Maximum Cup Depth=399 μm; RNFLT=110 μm; Scan Quality Score=32.6.

Figure 1A and 1B, respectively, show a vertically oriented B-scan through the ONH (orthogonal to the FoBMO axis) and a horizontally oriented B-scan (along the FoBMO axis). The pair of green line segments in each B-scan shown in Figs. 1A–1B represent the shortest distance from the BMO point to the ILM segmentation, that is, defining the parameter minimum rim width (MRW). Figure 1C shows projected 3D representations of the ILM and BMO for all 24 delineated B-scans viewed from the inferior sector of this example right eye from a P3 participant (with the temporal sector toward the left). The 48 MRW vectors (green line segments) are also shown. For this study we averaged these 48 MRW samples in each eye to obtain the global average MRW parameter. MRW Angle was defined as 0 degrees when a vector points inward on the BMO plane and as 90 degrees when it points anteriorly (upward in the image), perpendicular to the BMO plane; the average MRW Angle for each eye was used for this study.

The ONH minimum rim area (MRA) parameter was derived by summing the areas of each of the 48 trapezoids whose base is 7.5 degrees wide, centered on a Bruch’s membrane opening point, and whose height is the distance to the ILM segmentation that minimizes its area.¹⁵ The collection of these trapezoids is shown as the red, ribbon-like structure in Figure 1D also from a perspective looking toward the inferior pole of the ONH. MRA Angle is defined in the same way as MRW Angle. This eye has a very shallow MRW and MRA Angle (approaching 0 degrees) along the nasal rim.

The corresponding circumpapillary B-scan for this eye is shown in Figure 1E with segmentations for the ILM (yellow contour) and the posterior boundary of the retinal nerve fiber layer (RNFL, red contour). The parameter RNFL thickness (RNFLT) used in this study represents the global average thickness (of all 768 samples).

Additional parameters used in this study were derived from the three-dimensional (3D) fit of an ellipse (via minimization of RMS error) to the 48 delineated BMO points. BMO Area represents the area of that ellipse; BMO non-planarity was calculated as the standard deviation of axial distances between each BMO point and the fitted ellipse; BMO Aspect Ratio was calculated as the ratio of the long axis to the short axis of that ellipse. The parameter Mean Cup Depth represents the average axial depth of all ILM samples within the BMO, referenced to the top-of-the-rim (because that reference provides a measure of cup depth that is more independent of rim thickness as compared with using the BMO plane, Supplemental Figure 1). The parameter Maximum Cup Depth represents the most posterior sample of the ILM segmentation within the BMO; its value was referenced to the 3D BMO plane (such that a positive value indicates that none of the ILM samples were posterior to the BMO plane). In this case, the 3D-determined BMO plane provides a more precise estimate against which to reference the single most posterior voxel within the cup (versus the much larger non-planarity of the rim).

Grading ONH Prelaminar Schisis in OCT Scans

For GL/S eyes (in which OCT scans are obtained every six months) we graded the most recently available scans. In all eyes, the same OCT scans used for grading of ONH prelaminar schisis were also used for quantification of structural parameters. Two graders (EAL and BF) independently reviewed the ONH OCT scans (24-radial B-scans) for each eye in the study via the “Display” tab of the instrument’s user interface. Graders were masked to all patient demographic, clinical and ocular information, including visual field results and quantitative parameters from the OCT scans, as well as to each other’s grades. A four-point scale (zero to 3) was used for grading the severity of ONH schisis in each eye, which corresponded essentially to: 0=none; 1=mild; 2=moderate and 3=severe (see Figures 2 and 3). Specifically, the lateral boundary defining the “ONH” for the purpose of grading was the pair of BMO points within each B-scan; Grade-1 was applied when the size of the apparent optical void or voids were relatively small and located within the most anterior portion of the ONH prelaminar tissue (i.e., just beneath the surface). Grade-2 was applied to eyes in which the schisis was larger and/or located deeper within the prelaminar tissue (because schisis at deeper locations implies disruption of axo-glial tissue architecture). Grade-3 was applied to eyes in which the schisis volume was much larger in size (e.g., within a given B-scan and/or visible in many more B-scans) and associated with definite disruption of the sub-surface ONH prelaminar tissue (Figure 2). Graders also noted whether or not there was clear evidence of vitreous adhesions to the ONH surface (i.e., within the BMO boundary) and whether or not there was evidence of posterior vitreous cortex adhering to the peripapillary retina and/or that the peripapillary ILM was lifted away from underlying retina (see Figures 2 and 3). There was good agreement between the two graders initial scores, with identical scoring of 272 eyes (70.5%) and scores within one point in 373 eyes (96.6%). Overall, there was substantial agreement between the two graders for schisis versus no schisis (kappa=0.66, 95% confidence interval: 0.58 – 0.73) and for the specific schisis grade (weighted kappa=0.62). Among the 114/386 eyes for which there was an initial difference between graders, the vast majority (95/114, 83%) were in the GL/S group and most of those differed by only one level (87/95, 92%). OCT scan quality score (SQS) was not a significant influence on the initial inter-grader differences (p=0.180), probably because SQS was above 20 in all eyes (Table 1). Subsequently, all disagreements between graders were reconciled by consensus to achieve a final grade for each eye; graders remained masked to all other information during the consensus process.

Figure 2. Examples of ONH prelaminar schisis in GL/S eyes.

Each panel contains one of the ONH OCT radial B-scans of a GL/S eye from the P3 study group; the radial orientation of the B-scan is indicated by the bold green line in the inset at the lower right corner of the panel (as in Fig. 1). Red arrows point to optical voids whose size and depth were used to determine severity grade of ONH schisis. Green arrows point to locations where the peripapillary ILM appears to be separated from the underlying RNFL tissue. Open red arrows point to peripapillary retinoschisis.

Figure 3. Examples of ONH prelaminar schisis in eyes of healthy control group.

Each panel contains one of the ONH OCT radial B-scans of an eye from the healthy control group; the radial orientation of the B-scan is indicated by the bold green line in the inset at the lower right corner of the panel (as in Figs. 1 and 2). Red arrows point to optical voids whose size and depth were used to determine severity grade of ONH schisis. Green arrows point to locations where the peripapillary ILM appears to be separated from the underlying RNFL tissue. White arrows point to locations where the vitreous appears adherent to the ONH tissue.

Table 1:

Demographic and ocular characteristics of the two study groups: glaucoma/suspects and healthy controls.

	Glaucoma/Suspect 298 Eyes of 150 Persons		Healthy Controls 88 Eyes of 44 Persons		P-value
Parameter	Mean ± SD	Range	Mean ± SD	Range
Age (years)	72.2 ± 9.0	48.8 – 94.0	52.6 ± 17.4	22.3 – 84.4	<0.0001a
Gender	62 Male (41%) / 88 Female (59%)		19 Male (43%) / 25 Female (57%)		0.863b
Abnormal or Borderline GHT	113 (38%)		0 (0%)		<0.0001b
MD (dB)	−1.19 ± 3.65	−20.74 – 2.83	0.28 ± 0.87	−1.52 – 2.08	0.005c
PSD (dB)	2.95 ± 2.89	0.97 – 15.11	1.43 ± 0.26	0.94 – 2.65	<0.0001c
BMO Area (mm2)	2.00 ± 0.44	1.07 – 3.40	1.72 ± 0.31	1.15 – 2.45	<0.0001c
BMO non-planarity (pm)	22.2 ± 9.3	5.7 – 78.2	20.3 ± 8.7	8.1 – 67.5	0.053c
BMO Aspect Ratio	1.11 ± 0.06	1.01 – 1.32	1.14 ± 0.07	1.01 – 1.33	0.001c
MRW (pm)	240.8 ± 67.9	56.0 – 416.6	348.0 ± 63.4	236.1 – 549.3	<0.0001c
MRW Angle (degrees)	50.5 ± 15.5	11.7 – 81.6	64.6 ± 9.9	35.5 – 81.3	<0.0001c
MRA (mm2)	1.06 ± 0.28	0.27 – 1.89	1.36 ± 0.23	0.85 – 2.10	<0.0001c
MRA Angle (degrees)	37.0 ± 11.6	12.8 – 69.4	36.1 ± 10.8	19.4 – 71.0	0.398c
RNFLT (pm)	82.7 ± 17.0	27.8 – 131.3	96.7 ± 10.6	63.1 – 121.2	<0.0001c
Mean Cup Depth (pm)	314.6 ± 120.9	78.7 – 723.7	222.3 ± 84.6	23.9 – 450.1	<0.0001c
Max Cup Depth (pm)	334.5 ± 185.1	−137.5 – 891.5	171.5 ± 200.1	−263.6 – 613.3	<0.0001a
Scan Quality Score	30.2 ± 3.2	21.3 – 39.7	29.5 ± 3.4	20.8 – 38.4	0.042c
Vitreous-to-ONH	78 (26.2%)		43 (48.9%)		<0.0001b
Peripapillary ILM Lift	112 (37.6%)		67 (76.1%)		<0.0001b

Standard automated perimetry visual field parameters: glaucoma hemifield test (GHT); mean deviation (MD); pattern standard deviation (PSD); optical coherence tomography (OCT) parameters: Bruch’s membrane opening (BMO); minimum rim width (MRW); minimum rim area (MRA); retinal nerve fiber layer thickness (RNFLT); optic nerve head (ONH); internal limiting membrane (ILM).

^aWelch’s t-test

^bFisher’s exact test

^cMann Whitney t-test

Analysis and Statistics

Statistical analysis was performed using either a commercial software package (Prism 8, GraphPad Software, Inc., La Jolla, CA) or the R language and environment for statistical computing (R version 3.3.3, R Core Team, Vienna, 2013, http://www.R-project.org/). Differences between study groups were evaluated using either an unpaired, two-sided t-test with Welch’s correction for continuous variables such as visual field indices and OCT parameters or by Fisher’s exact test for binary variables. To evaluate associations between ONH schisis and each of the potential predictors, two sets of analyses were performed. First, logistic models were used to predict the odds of the detection of any ONH schisis (i.e. scores “1”, “2” or “3” vs. score “0”), using a generalized estimating equation framework to account for the presence of fellow eyes in the dataset. Second, proportional odds ordinal logistic regression models were used to predict the actual score (i.e. making the implicit assumption that the effect of a predictor on the odds of being score “1” vs. score “0” exactly equals the effect on the odds of being score “2” vs. score “1”, or of being score “3” vs. score “2”); using robust sandwich estimates of the standard errors to account for the presence of fellow eyes. In each case, various potential predictors of ONH schisis were evaluated univariately. Predictors that had a univariate p<0.2 were then entered into a multivariable model, and a final model was derived by single backward elimination. All other statistical analyses are listed in context with presentation of corresponding results in the next section.

Results

Table 1 lists the demographic characteristics, visual field and OCT structural parameters measured in the two study groups. As expected, there were substantial differences between the GL/S group and the healthy control group for parameters reflecting glaucomatous visual field and structural damage: visual field indices MD and PSD and the frequency of GHT being abnormal, as well as the OCT parameters MRW, MRW angle, MRA, Mean Cup Depth, Maximum Cup Depth and circumpapillary RNFL thickness (all p<0.0001). In addition, the OCT parameter BMO Area was 16% larger in GL/S eyes, on average, than in healthy control eyes (p<0.0001); BMO Aspect Ratio was slightly smaller (i.e., BMO was rounder, or less oval, p=0.001) and the BMO tended toward greater non-planarity (p=0.08) in GL/S eyes. The frequency of observing either vitreous adhesions to the ONH surface or separation (“lift”) of the peripapillary ILM was higher in healthy control eyes (each p<0.0001). Although participants in the GL/S group were 20 years older, on average, than the healthy controls (p<0.0001), their OCT Scan Quality Score tended to be slightly better (p=0.07).

The frequency and severity of ONH prelaminar schisis were greater in GL/S eyes than in healthy controls (chi-square=11.54, p=0.009, Figure 4). Among the GL/S group, 165 eyes (55.4%) had no visible schisis (Grade-0), 71 (23.8%) had Grade-1,46 (15.4%) had Grade-2 and 16 (5.4%) had Grade-3 schisis. Among the healthy control eyes, 59 (67.0%) had Grade-0, 24 (27.3%) had Grade-1, 5 (5.7%) had Grade-2; none of the healthy control eyes had Grade-3 schisis. The optical voids defining the OCT appearance of ONH schisis were generally larger and located deeper into the prelaminar ONH tissue, resulting in a more disrupted appearance of the tissue architecture in the GL/S eyes as compared to healthy control eyes (e.g., Figs. 2 and 3).

Figure 4. Frequency and severity of ONH prelaminar schisis in each study group.

Stacked frequency histogram shows that ONH prelaminar schisis was observed more frequently and had higher severity grades among the GL/S group as compared to the healthy control group (chi-square=11.54, p=0.009; color indicates schisis grade).

Figure 5 shows the distribution of visual field and OCT parameters by study group and ONH schisis grade. The array of plots in Fig. 5 reveal several parameters that appear to have a strong association with ONH schisis severity. For example, ONH schisis grade exhibits a clear association with MRW (Fig. 5G), MRW Angle (Fig. 5H), and Mean Cup Depth (Fig. 5I), driven largely by the group of GL/S eyes, while the association with other parameters such as BMO Area (Fig. 5D) and Maximum Cup Depth (Fig. 5L) appears to manifest in both GL/S and Healthy Control eyes. In contrast, while some parameters such as age and circumpapillary RNFL thickness show a clear difference between the GL/S and Healthy Control groups, there is not much association with ONH schisis apparent in either group. Specifically, there was no significant difference in age between control eyes grouped by ONH schisis grade (Grade-0, 1 or 2; p=0.483 one-way ANOVA), nor between the four groups of GL/S eyes (Grade-0, 1, 2 or 3; p=0.300).

Figure 5. Distribution of visual field and OCT parameters by study group and ONH schisis grade.

Box plots represent the median, inter-quartile range and extremes of each distribution; healthy control group (HC) shown in blue and glaucoma/suspect (GL/S) group shown in red.

To formally evaluate the association between ONH prelaminar schisis and each of these potential predictors, we used two types of logistic models (as detailed above in Methods). Table 2 shows the effect of each potential predictor in two types of univariate analyses. The first set of logistic models were used to predict the odds of detecting any ONH schisis (i.e. score of “1”, “2” or “3” vs. score of “0”); the second set were proportional odds ordinal logistic regression models used to predict the actual schisis score. The two types of analysis returned very similar results, consistent with the findings presented in Fig. 5. The frequency and severity grade of ONH prelaminar schisis exhibited significant associations with several common clinical measures of glaucomatous damage; OCT parameters reflecting glaucomatous deformation of the ONH, in particular, MRW, MRW Angle, Mean and Maximum Cup Depth appear to be the most robust predictors of ONH schisis being present and being more severe.

Table 2: The effect of each predictor in univariate analyses.

The second column shows the increment of that predictor used; in each case the increment is chosen to give the odds ratio for if an eye has a clinically-relevant magnitude of worse glaucomatous damage. The third column shows the odds ratio for the presence of schisis (i.e. scores “1”, “2” or “3”) compared to no schisis (i.e. score “0”) when the predictor changes by the stated increment, in a generalized estimating equation logistic regression model (with the associated 95% confidence interval); the fourth column shows the significance of this odds ratio. The fifth column shows the odds ratio from a proportional odds model treating score as an ordered factor variable (together with its confidence interval); the final column shows the significance of this odds ratio.

Predictor	Increment	Odds Ratio	p-value	Odds Ratio	p-value
Group	In the P3 group	1.64 (0.90–2.98)	0.10	1.88 (1.17–3.08)	0.02
Age	10 years older	1.16 (0.98–1.37)	0.09	1.15 (1.00–1.33)	0.09
Abnormal GHT	Abnormal or Borderline	1.80 (1.13–2.88)	0.01	1.87 (1.23 – 2.85)	0.009
MD	1 dB more negative	1.07 (0.99–1.15)	0.07	1.07 (1.02–1.14)	0.03
PSD	1 dB more positive	1.12 (1.01–1.23)	0.03	1.12 (1.05–1.20)	0.008
BMO Area	0.1 mm2 larger	1.07 (1.02–1.14)	0.01	1.07 (1.03–1.12)	0.005
BMO non-planarity	1 pm smaller	1.03 (0.99–1.04)	0.19	1.01 (0.99–1.04)	0.26
BMO Aspect Ratio	0.1 units smaller	1.23 (0.85–1.77)	0.27	1.26 (0.92–1.75)	0.17
MRW	10 pm smaller	1.06 (1.03–1.09)	0.0001	1.07 (1.04–1.10)	<0.0001
MRW Angle	1° smaller	1.04 (1.02–1.05)	<0.0001	1.04 (1.03–1.05)	<0.0001
MRA	0.1 mm2 smaller	1.10 (1.02–1.19)	0.02	1.13 (1.05–1.21)	0.003
MRA Angle	1° smaller	1.02 (1.00–1.04)	0.09	1.02 (1.00–1.04)	0.06
RNFLT	1 pm smaller	1.01 (1.00–1.03)	0.17	1.01 (1.00–1.02)	0.11
Mean Cup Depth	10 pm larger	1.04 (1.02–1.06)	0.0003	1.04 (1.02–1.06)	<0.0001
Max Cup Depth	10 pm larger	1.03 (1.02–1.04)	<0.0001	1.03 (1.02–1.04)	<0.0001
Quality Score	1 unit lower	0.99 (0.93–1.06)	0.79	0.98 (0.92–1.04)	0.57
Vitreous-to-ONH	True	1.17 (0.72–1.90)	0.52	1.17 (0.77–1.77)	0.52
Peripapillary ILM Lift	True	0.91 (0.57–1.47)	0.71	0.99 (0.66–1.46)	0.95

Starting with the variables that achieved p<0.2 in the univariate models shown in the final column of Table 2, we created a multivariable proportional odds ordered logistic model to predict ONH schisis grade. In order to minimize effects of instability due to collinearity, MRW and MRW Angle were included but MRA and MRA Angle omitted (since the MRW parameters showed slightly stronger relations to schisis than the MRA parameters, likely due to MRW being less variable than MRA); and with Maximum Cup Depth used but Mean Cup Depth omitted (by the same logic). After single backward elimination, the final model predicted that ONH schisis grade would be higher in eyes with larger BMO Area (odds ratio 1.06 for 0.1mm² larger BMO area, 95% confidence interval 1.01–1.11, p=0.028) and smaller MRW Angle (odds ratio 1.04 for 1° smaller angle, 95% confidence interval 1.02–1.05, p<0.0001); these tend to result in a clinical appearance of having a larger and deeper cup, respectively.

While age was not a significant predictor of ONH schisis presence or severity in our multivariable analyses after backwards elimination, we conducted further analysis retaining age in the multivariable model because the average age of the GL/S and Healthy Control groups differed by 19.6 years (Table 1) and because this age range is commonly when interactions between the vitreous and the ILM can change (such as occurring during posterior vitreous detachment). When we repeated the multivariable analysis retaining age as a predictor in all models, rather than eliminating it via the usual backwards elimination process, ONH schisis grade was still predicted by greater BMO area (p=0.031) and smaller MRW angle (p<0.001), but not by age (p=0.456). Similarly, when using MRA instead of MRW, age was not significant in the final model (p=0.882).

Discussion

Here we report that ONH prelaminar schisis is more commonly observed in OCT scans of eyes with glaucoma, and that its presence and severity are associated with the severity of glaucomatous damage as measured using a variety of standard clinical diagnostic parameters. This sign should not be considered diagnostic in isolation as mild examples were also found in the healthy control eyes. However, in the vast majority of the healthy eyes, the optical void was small and located just below to the surface, appearing to be separation of the inner limiting membrane of Elschnig and the central meniscus of Kuhnt¹⁶ away from the underlying ONH tissue in a highly localized area (e.g., Fig. 3). In contrast, in the GL/S eyes, the optical voids were generally much larger and often located deeper within the sub-surface ONH tissue. In the most severe examples, the ONH schisis was associated with clear disruption of the deeper prelaminar ONH tissue and in some cases the tissue disruption appeared to include the lamina cribrosa (e.g., Fig. 2G) and peripapillary retina (e.g., Fig. 2H). Such examples were never observed in healthy control eyes. These observations may be related to lamina cribrosa defects in glaucoma detected by OCT,^17–19 but are distinctly different insofar as the tissue disruption is within the prelaminar ONH. Isolated case reports or small series have shown similar findings in eyes with optic pit, coloboma or other cavitary ONH abnormalities,^{20, 21} including one with coexisting glaucoma,²² and other examples in glaucoma,²³ but ours is the first systematic study to document the frequency and severity spectrum of this finding in a larger cohort and demonstrate an association with the degree of glaucomatous ONH deformation.

Thus, the presence of a more severe ONH prelaminar schisis should alert clinicians to a higher risk of ongoing glaucomatous damage, as supported by the clear associations we report here for common structural parameters such as ONH size (BMO Area), neuro-retinal rim width and cup depth. Although it is not used commonly for clinical diagnostics, the MRW Angle parameter also exhibited a strong association with ONH prelaminar schisis and has previously been shown to track closely with ONH deformation in experimental glaucoma.²⁴ While prelaminar schisis might reflect ongoing structural change due to glaucomatous deformation and remodeling of the ONH, likely due to mechanical failure of the load-bearing connective tissues, it is also possible that it represents a risk factor for further damage. Future longitudinal studies are needed to determine if ONH prelaminar schisis is itself predictive of more rapid subsequent glaucoma progression.

Another important clinical implication for ONH prelaminar schisis is its potential to influence diagnostic parameters derived from OCT scans. Figure 6 presents another case example of a Grade-3 ONH schisis in one of the GL/S eyes. The purpose of providing this particular example is three-fold. First, it demonstrates that ONH schisis can impact the diagnostic parameters and the clinical report produced by the instrument – in this particular case by creating risk of “green disease” whereby the ONH rim thickness (MRW parameter) is reported to be within normal limits for the superior pole despite the true rim thickness being extremely narrow (~ 50 μm, not 250 μm as reported). Thus, it also serves to reiterate the importance of routinely inspecting the actual B-scans and their segmentations in order to minimize risk of misdiagnosis.¹⁴ Second, it underscores the need for scientists and industry to develop more elaborate and robust segmentation capabilities to account for unusual anatomy, such as exists with ONH prelaminar schisis. Finally, this example also suggests that the strength of cup depth as a predictor of ONH prelaminar schisis might be even greater than we report here because the true cup depth in some cases with higher grade ONH schisis might be even larger, given that the ILM segmentation generally does not account for the schisis space. In order to help visualize this scenario, Supplemental Figure 2 shows the ILM and BMO segmentations for the same 8 examples shown earlier in Fig. 2.

Figure 6. Example showing impact of ONH prelaminar schisis on clinical diagnostic parameters.

The image shows the clinical instrument report for the MRW parameter in an eye from the GL/S group with a Grade-3 ONH schisis (male, age: 87 years; MD: −6.1 dB; PSD: 9.3 dB; RNFLT: 61.0 μm; BMO Area 2.01 mm²; MRW: 221 μm; Mean Cup Depth: 340 μm). A potentially important clinical consequence of ONH schisis is its influence on OCT image segmentation and diagnostic parameters. This example demonstrates how the true MRW values can be dramatically overestimated and cup depth values underestimated. A corollary of this is that our results probably underestimate the association between these parameters and the presence of ONH schisis.

ONH prelaminar schisis may also have implications for glaucoma pathophysiology. For example, we previously proposed that ONH prelaminar schisis may include detachment of the blood vessels and aspects of their adventitia from the glio-fibrillar meniscus of Kuhnt, the ILM of Elschnig, and in some cases also from underlying prelaminar tissue as posterior deformation of the ONH progresses – because the blood vessels are tethered throughout the retina and might thus be limited in how far posteriorly they can be displaced as “cupping” progresses.¹⁰ We surmised this initially from scrutiny of structural OCT scans. However, Figure 7 shows an OCT-angiography scan from an eye with prominent signs of ONH prelaminar schisis. This example clearly demonstrates that major blood vessels are sometimes contained within the strands of tissue separated anteriorly from the underlying prelaminar tissue, a form of “vessel baring” or “bridging” within the ONH rim. In contrast, the major blood vessels are usually embedded within the rim tissue in healthy eyes. Such changes may also have a role in the development of optic disc (Drance) hemorrhages in glaucoma, a topic of study currently underway. Further support for this scenario derives from the observation that major blood vessel cross-sections within the ONH rim tissue are rarely ever observed posterior to the BMO in radially-oriented OCT B-scans of healthy eyes. In contrast, this is quite commonly observed in eyes with glaucomatous deformation (e.g., Fig. 2G). Changes in vessel position, both within the ONH^{25, 26} and the peripapillary retina²⁷ are a well-known sign of glaucomatous deformation and are associated with disease progression. Longitudinal tracking of vessel position may also help determine relationships between glaucomatous ONH changes and development of prelaminar schisis, retinal abnormalities like peripapillary retinoschisis and paravascular defects.¹⁰

Figure 7. Example of OCT-Angiography demonstrating how the vasculature is involved in some cases of ONH schisis.

In this eye (the same eye shown in Fig. 2H), major blood vessel trunks are separated anteriorly from the underlying prelaminar tissue (arrow in C); there are strands of presumably connective tissue spanning the optic cup, which appear connected to this vessel.

In addition to the mechanical consequences of glaucomatous ONH deformation and remodeling, there may well also be alterations of physiology contributing to failure of homeostasis within the tissue under stress. For example, aquaporin-4 (AQP4) is a water channel thought to be primarily responsible for maintenance of water flux in the retina and optic nerve as evidenced by its highly polarized distribution in Müller glia and astrocytes.²⁸ Yet it is not expressed by the astrocytes within the prelaminar portion of the ONH in rodents, whereas an entirely different aquaporin channel (AQP9) is instead.^{28, 29} This suggests some potentially important differences in the way water flux is handled in this highly specialized transition region. Moreover, ONH astrocytes are known to respond rapidly to IOP elevation by retracting and reorienting their processes,^30–34 as well as by exhibiting altered expression of aquaporin channels.^{29, 35} These changes may be beneficial for some aspects of the response, but the presumed loss of the astrocyte syncytium functions, such as those dependent on end-foot contacts where aquaporin and rectifying potassium channels are normally enriched, might result in reduced capacity to handle ionic and water flux, potentially leading to fluid accumulations, among other physiological disturbances. This might even further exacerbate mechanical challenges faced by the ONH tissues during glaucomatous deformation and remodeling.

Limitations of this study include that the population is nearly entirely Caucasian, which may limit its generalizability. Moreover, the grading scale we used for evaluation of ONH prelaminar schisis was subjective and mostly qualitative. However, we found generally good agreement between graders and reduced the risk of bias by using two independent masked observers and a combination of logistic regression approaches for analysis, which produced similar results, thus supporting their robustness. Nevertheless, future studies might benefit from more advanced image segmentation techniques and/or enhanced contrast provided by other imaging modalities, perhaps aided by artificial intelligence, in order to detect and quantify aspect of ONH schisis objectively. In a similar context, our study used the ONH-RC scan pattern of 24 radials; dense raster scans might increase detection of ONH schisis and improve quantification of size. Moreover, the radial scan pattern samples the central portion of the ONH more densely than the lateral portions where blood vessel shadows more commonly obscure underlying details. Graders tried to account for position within the ONH when considering the size or extent of voids representing ONH schisis, but this was likely one important factor contributing to initial differences between graders.

Additional limitations include that the GL/S group includes relatively few GL eyes with more advanced damage and that the healthy control group was a relatively small sample of only 88 eyes of 44 individuals. This was smaller than the GL/S group and also 20 years younger, on average. However, age was not a significant predictor of ONH prelaminar schisis in a univariate model, nor was age a significant predictor in any of the multivariable models in which it was included, so it is unlikely that the age difference mattered in this regard. While our results do show a significant association between glaucoma severity and the presence or grade of ONH schisis, this was much stronger for the structural parameters than for visual field indices. For example, the visual field indices were not significant predictors in multivariable models that also included structural predictors, suggesting that glaucomatous structural damage is more strongly associated with ONH schisis (but functional damage less so), or that the measures of function are largely redundant with those of structure. However, only 24 of the 298 GL/S eyes had MD worse than −6 dB, which may be too few to more fully evaluate the relationship between ONH schisis and stage of glaucomatous functional loss. Further work to evaluate larger populations of healthy, glaucomatous and suspect eyes with more varied ethnicity and a wider range of glaucomatous visual field loss will help to refine and extend our understanding of how ONH schisis develops and to what extent it is related to the stage of glaucomatous damage. Moreover, future longitudinal studies will help determine whether ONH schisis is a consequence of glaucoma progression and/or itself a predictor of future glaucoma progression, as well as whether its appearance changes over time.

In summary, we found that the ONH prelaminar schisis sign was detected more frequently in OCT scans of glaucomatous eyes compared to healthy control eyes and that its presence and severity are associated with common OCT metrics related to increased risk, ONH deformation and structural damage in glaucoma, such as larger ONH size, thinner neuro-retinal rim width and greater cup depth.