Gaze Evoked Deformations of the Peripapillary Retina in Papilledema and Ischemic Optic Neuropathy

Abstract

Purpose: To examine the effects of horizontal eye movements on the shape of the peripapillary basement membrane layer (ppBM layer) in patients with papilledema, anterior ischemic optic neuropathy (AION) and normal eyes.

Methods: Spectral-domain optical coherence tomography (SD-OCT) axial rasters of the optic nerve were used to analyze the shape of the ppBM layer. We compared registered images in two eye positions: 10° to 15° of adduction and 30° to 40° of abduction from 80 patients (45 with papilledema, 15 with AION, and 20 normal eyes).

Results: Horizontal eye movements induce a relative “seesaw-like” shape deformation of the ppBM layer in patients with papilledema. On adduction, there is a relative posterior displacement temporal to the basement membrane opening (BMO) and an anterior displacement nasally (P = 0.001). The pattern reverses in abduction. Normal eyes and AION display similar but smaller changes in shape predominantly affecting the temporal side on adduction. The shape difference was significantly different for normal eyes (P = 0.03) and for AION (P = 0.01).

Conclusions: Horizontal eye movements affect the shape of the ppBM layer. The deformation in normals and AION, in adduction, causes posterior displacement temporal to the BMO. In contrast, the deformations in papilledema are larger, involving temporal and nasal sides, presumably because of shifts in cerebrospinal fluid pressure against the scleral flange and hydraulic stiffening of the optic nerve sheath. The clinical importance of gaze-induced deformations is unknown but repetitive motion may be a factor in the genesis or progression of a variety of optic neuropathies.

The scleral canal opening, through which the optic nerve emerges from the eye, is a mechanically stressed environment.¹ Elevated fluid pressures (in the intraocular or cerebrospinal fluid [CSF] compartments) concentrate opposing loads on the lamina cribrosa and peripapillary sclera that may, depending on the magnitude and direction of the pressure gradient, cause glaucomatous optic atrophy or papilledema. The intraorbital optic nerve, particularly at the scleral canal, also is stressed by near continuous movements of the eye that whipsaws the nerve with each change in eye position. 

The idea that gaze might strain the optic nerve and surrounding eye wall dates back to Purkinje in the early 19th century, who suggested that traction on the optic nerve might explain certain gaze-evoked phosphenes.^2–7 In 1941, Friedman⁴ speculated that on adduction the peripapillary retina temporal to the disc was tethered by the optic nerve and compressed by the sclera on the nasal side. In 1980, Greene¹ examined stress concentration on the posterior sclera caused by the extraocular muscles in myopia. A number of investigators have shown gaze-evoked changes in intraocular pressure and axial length presumably mediated by the effects of the eye muscles on the globe.^8,9 Imaging studies examining the position of the optic nerve during ocular rotation also have been described.^10–12 These studies provide compelling evidence that eye movements can biomechanically stress and strain the optic nerve and or surrounding eye wall. 

In 2008, we speculated that eye movements may have a role in the genesis of spontaneously acquired asymptomatic, peripapillary subretinal hemorrhages in patients with crowded tilted discs and myopia who otherwise were normal.¹³ Since then, I have observed that gaze-induced shape changes of the peripapillary basement membrane layer (ppBM layer) can occur in papilledema and that this disparity occasionally can provide useful clinical information. For the last several years we have incorporated axial rasters obtained in eccentric gaze as part of our spectral-domain optical coherence tomography (SD-OCT) evaluation in selected patients. Using a shape analysis technique (geometric morphometrics [GM] we have shown previously that the SD-OCT is capable of imaging and quantifying peripapillary deformations of the eye wall caused by changes in intracranial pressure.^14,15 This report extends these observations by retrospectively investigating the effects of eccentric eye position on the shape of the peripapillary eye wall using axial rasters of the SD-OCT. To distinguish the relative contributions, if any, of intracranial pressure and optic disc edema, we studied three groups of patients: idiopathic intracranial hypertension with papilledema, nonarteritic anterior ischemic optic neuropathy (with similar degrees of disc edema), and normal eyes. 

Methods

Subjects

We retrospectively reviewed our database of SD-OCT images from 80 patients who met the inclusion criteria defined below. They included 45 patients with new onset idiopathic intracranial hypertension with papilledema (mean age of 31 ± 8 years, 41 women, 4 men), 15 with acute nonarteritic anterior ischemic optic neuropathy (AION; mean age of 60 ± 12, 11 men, 4 women), and 20 subjects with normal eyes (average age of 43 ± 15, 8 men, 12 women). 

Papilledema.

All subjects were newly diagnosed cases that met the modified Dandy criteria¹⁶ for the diagnosis of idiopathic intracranial hypertension (IIH) with opening pressures of >25 cm H₂O and papilledema. The diagnosis of papilledema was based on funduscopic features that included obscuration of the disc margins, elevation, engorgement of the vasculature, and in some cases retinal and choroidal folds and hemorrhages. By SD-OCT, all had an abnormal (>95% of the normal controls) mean retinal nerve fiber layer (RNFL) thickness. 

Nonarteritic Anterior Ischemic Optic Neuropathy (AION).

This group included patients over the age of 40 with signs of an optic neuropathy characterized by the acute, recent (<2 weeks), painless onset of unilateral vision loss, a visual field defect on Humphrey perimeter, a relative afferent pupillary defect, and a swollen optic disc associated with thickening of the mean RNFL (>95% of the normal controls). Patients were excluded if they had painful eye movements, a history of any other neurologic or systemic disease other than hypertension, diabetes, atherosclerosis, and hyperlipidemia. If there was any question about the diagnosis, the patient also underwent magnetic resonance imaging to rule out a compressive optic neuropathy or demyelinating disease. Any patient with a diagnosis of multiple sclerosis or intracranial lesions was excluded. 

Normal controls were recruited from patients who presented for routine examinations and had a completely normal ophthalmic examination with an unremarkable review of systems (specifically including questions about headaches, diplopia, and pulsatile tinnitus). Subjects with abnormalities in their corrected visual acuity, pupil exam, Ishihara color plate testing, intraocular pressure, Humphrey visual fields, or ophthalmoscopic findings were excluded. Spectral-domain OCT was performed and patients were excluded if there was evidence retinal nerve layer thinning, cupping, or edema. Any patient with a disc anomaly (e.g., tilted optic disc, high myopia, staphylomas, optic disc drusen, hypoplasia, or otherwise dysplastic) also was excluded. 

Image Acquisition

A Cirrus 5000 SD-OCT (Carl Zeiss Meditec, Inc., Dublin, CA, USA) was used to acquire images with a signal strength of ≥7. Two protocols imaging the optic nerve head were used: (1) optic disc cube 200 × 200 and (2) a 5-line, high definition, axial raster (9 mm in length, 0.125-mm intervals). The optic disc cube was used to calculate the mean RNFL thickness derived from 256 A-scan samples along the path of a calculation circle that was 3.46 mm in diameter centered on the optic disc. The most centrally positioned of the 5 lines in the raster scan of the optic disc was used for image analysis. This SD-OCT produces raster images with an aspect ratio of 3:2 (750 × 500 pixels [px]) which magnifies the vertical-axis to better visualize the layers of the retina (Fig. 1). We converted the aspect ratio from 3:2 to a true aspect ratio of 9:2 (750 × 167 px) for the geometric morphometric analysis of shape. For the purposes of demonstrating small differences, some of the Figures are displayed in a 3:2 aspect ratio where indicated. 

Figure 1

 

Spectral-domain OCT, 9 mm, axial raster of a patient with papilledema displayed with an aspect ratio of 3:2 (a) and 9:2 (b). The default display format for the Cirrus SD-OCT (3:2 aspect ratio), magnifies the image vertically to enhance viewing of the retinal layers, whereas the 9:2 aspect ratio images use a uniform vertical and horizontal spatial scale. The placement of 20 equidistant semilandmarks (red numbered circles) spanning a distance of 2500 μm temporally (points 1–10) and 2500 μm nasally (points 11–20) from the border of the BMO is shown. This image also demonstrates the characteristic anterior displacement (toward the vitreous) of the peripapillary BM layer in papilledema.

High definition 5 line axial rasters were obtained in two head positions with the eyes in adduction and abduction to the extent that the patient and device would permit. Adduction was limited by the patient's nose against the lens element of the device at approximately 10° to 15° and abduction was less restricted at approximately 30° to 40°. These ranges were based on rotational head positions while the patient viewed the OCT fixation target using a head mounted protractor in a sample set of subjects. The Zeiss Cirrus 5000 OCT uses a scanning laser ophthalmoscopic image stabilization process (“FastTrac”) to register images. The device tracks and also registers images (“Tracked to prior” option) using retinal features, such as blood vessel crossings, to ensure proper alignment and positioning of subsequent images. We used the most centrally positioned raster of the 5 raster lines in adduction as a baseline reference to compare to the axial raster obtained in abduction at the same location. 

The right eye was used in all subjects with papilledema and normals. If the left eye was affected in subjects with AION, the image was flipped horizontally to align the temporal and nasal regions across all subjects. 

GM Shape Analysis

Geometric Morphometrics was used to analyze the shape of ppBM layer imaged on the SD-OCT raster. This is a well-established method that quantifies and analyzes shape and its covariation with other variables.^17–19 The details of this methodology go beyond the scope of this paper; however, its application to the analysis of the optic nerve head has been described previously.^14,15,18,20 Only a brief summary is presented below. The interested reader can obtain details in the monograph by Zelditch et al.¹⁷ For a comprehensive resource on GM including online software please see Morphometrics at SUNY Stony Brook, (available in the public domain; http://life.bio.sunysb.edu/morph/).¹⁹ In GM, shape is that geometric property that remains after normalizing for differences in centroid location, scale, and rotation. The analysis is accomplished using a variety of software applications.^{17–19,21,22} The “tps” software suite¹⁹ was used in this shape analysis. 

Digitizing Semilandmarks

Using imaging software (Photoshop; Adobe Systems, San Jose, CA, USA) we superimposed a 2500 μ rectilinear grid on the temporal and nasal side of the basement membrane opening (BMO). Using the true (9:2) aspect ratio images, a grid was used to locate 10 equidistant “semilandmarks” (landmarks placed along a curve or surface) on both sides of the BMO. Each semilandmark was positioned along the outer edge of the retinal pigment epithelial–basement membrane layer (BM layer) starting at the edge of the BMO (Fig. 1). Semilandmark placement was performed on image files that were deidentified with respect to name and eye position. 

The ppBM layer was used because it is easily identified in nearly all eyes, including those with severe optic disc edema that can sometimes shadow subsurface structures. If the ppBM layer was not clearly visualized, the images were excluded. This layer may not always reflect the overall shape of the eye wall particularly in the presence of choroidal or subretinal effusions; however, previous studies have shown that it can serve as a reasonably accurate surrogate for the overall subsurface shape of the globe at that location.^14,15,23,24 

Generalized Procrustes Analysis (GPA)

Generalized Procrustes analysis, or Procrustes superimposition, is the process of superimposing all of the specimen shapes onto a mean shape in three sequential steps.¹⁷ First, the centroid of the semilandmarks is translated to the origin. Second, each shape is rescaled to a uniform size. Third, rotational differences are minimized between corresponding landmarks. 

Thin Plate Spline

The thin plate spline displays differences in shape as a smooth deformation using an algorithm that interpolates differences between landmarks that can be visualized using vectors at each landmark. The thin plate spline also defines a set of shape variables (partial warps) that capture the differences between shapes. The partial warp scores generate data matrices with the proper degrees of freedom that can be analyzed using multivariate statistical methods. 

Principal Component Analysis (PCA)

Principle component analysis is used to express and ordinate the variation in shape along a series of dimensions that are linear combinations of the partial warps. In effect, PCA identifies the critical components of shape variance, thereby simplifying the dimensional complexity to determine if there are distinctive patterns associated with a specific cohort (in this case a shape difference between abduction and adduction). The relative contribution of each dimension (principal component) is expressed as a percent of the total variance. Principal component analysis was performed on each cohort using a variance–covariance matrix of the shape variables.¹⁷ 

Each mark along the x and y axes of the PC plot represents 0.1000 “Procrustes distance units,” defined as the square root of the sum of squared differences between the positions of the landmarks from two superimposed shapes that have been normalized for centroid location, size, and rotation. These units represent shape difference expressed in terms of their respective principal component. 

Interpretation of Shape Changes and Displacement

One of the difficulties in estimating absolute displacements using the SD-OCT, particularly in patients with intracranial hypertension where deformations can be relatively large, is that it is not possible to establish a fixed reference plane inside the 9-mm window of the raster image. It is, however, possible to estimate the relative displacement. In this study we used the outermost nasal and temporal locations from the BMO along the BM layer as the reference plane for comparisons between abduction and adduction. These points were chosen for comparative illustrations on the assumption that maximal displacement occurs at the BMO and minimal at distant points. The shape analysis technique itself, however, does not depend on establishing an absolute reference plane. 

The Procrustes superimposition by definition normalizes for differences in size, location, and rotation between specimens and minimizes the remaining differences in shape between landmarks. Thus, the interpretation of shape in GM must consider the shape as a whole and not in terms of a single landmark. 

Lastly, it is important to distinguish between the “true” aspect ratio (9:2) and the “magnified” aspect ratio (3:2; see Fig. 1) in the interpretation of Figures. Because the observed differences between shapes, particularly among normals and AION, was small, we used a magnified aspect ratio to illustrate the differences in some of the Figures. Each Figure that follows will indicate if the images were magnified or true. However, for the purposes of GM shape analysis we used the true aspect ratio images to determine if there was a statistically significant difference in shape between abduction and adduction. 

Statistical Analysis of Shape

Statistical analysis of shape is based on comparing the sums of squared Procrustes differences between and within the samples expressed as an F-ratio.^17,25 The evaluation compares the observed F-value to a distribution based on 10,000 random permutations of the individuals to the groups being compared. The proportion of Goodall's F statistics from these permutations equal to or larger than the observed Goodall's statistic is interpreted as the “P value” for the test. The resampling method used in the permutation statistics in this study treats each eye (each shape vector) as independent – not the individual points used to capture the overall shape. To avoid correlation bias between eyes within an individual, all statistical analyses on shape were performed on the right eye or in unilateral cases (e.g., AION) the affected eye. A Students t-test was used to compare the means of continuous variables. An ANOVA was used to complete multiple comparisons of continuous variables. 

This study was approved and complied with policies of the SUNY Stony Brook Committee on Research Involving Humans and with Declaration of Helsinki. 

Results

We identified 80 patients who met the criteria defined above. The 45 patients with papilledema had a mean RNFL of 203 ± 87 μm; 15 patients with AION had a mean RNFL of 208 ± 67 μm. A t-test failed show a significant difference in the mean RNFL between AION and papilledema. Normal subjects had a mean RNFL of 94 ± 12 μm, which was significantly different from the two groups with disc edema (P < 0.001, ANOVA). 

Figure 2 shows the mean shape (“consensus” shape) for each group of patients in abduction and adduction. The shapes are vertically magnified (i.e., 3:2 aspect ratio) to display the small shape changes that occur in normals and AION. Irrespective of eye position, the shape of the ppBM layer in patients with papilledema was anteriorly displaced toward the vitreous, whereas normals and AION had a V-shaped BM layer deflected away from the vitreous. The differences between abduction and adduction are color encoded in each row of line tracings. In papilledema, there was an alternating tilt or “seesaw”-like difference in the ppBM layer. On adduction (black line tracing) the temporal region was relatively displaced posteriorly, whereas the nasal region was displaced anteriorly; the reverse pattern was seen in abduction (red line). The difference in shape between abduction and adduction in papilledema was statistically significant (permutation, P = 0.001). 

Figure 2

 

Summary of the consensus (mean) shapes of the ppBM layer in abduction (red line) and adduction (black line) for each group as labeled. The shapes in this Figure are displayed in a 3:2 aspect ratio to demonstrate the small gaze-evoked shape disparities in AION and normals. Irrespective of eye position, the shape of the ppBM layer in papilledema is displaced anteriorly, whereas normals and AION have a mild V-shape oriented posteriorly. In papilledema, there is a seesaw alternating tilt in the shape of the ppBM layer between adduction and abduction. A similar pattern can be seen in normals and AION, but the difference predominantly affects the ppBM layer temporal to the BMO.

In normal eyes there was a similar pattern predominantly affecting the temporal side. That is, on adduction there was a posterior displacement temporally and only mild anterior displacement nasally; again reversed on abduction. In AION the pattern was repeated temporally but with little difference on the nasal side. For both groups the difference between abduction and adduction was statistically significant at P = 0.03 (permutation) for normals and P = 0.01 (permutation) for AION. 

The difference between abduction and adduction in normals and AION was smaller in magnitude than in papilledema. In contrast to the relative symmetrical deformations on both sides of the BMO in papilledema, the changes in normals and AION appeared to be skewed temporally. That is, the temporal side relative to the nasal appeared to move posteriorly on adduction and anteriorly on abduction with little or no change on the nasal side. 

A variance–covariance matrix of the shape variables derived from the semilandmark data from all three groups of patients (Fig. 3) was used to perform a PCA (using tpsRelW software).¹⁹ The first three principal components accounted for 93% of the variance in shape; PC1, accounting for 78% of the variance, implies a shape characterized by an anterior–posterior axial displacement of the peripapillary BM layer. In other words, PC1 quantifies the degree to which the globe is indented or flattened toward (anteriorly, negative abscissa) or away from (posteriorly, positive abscissa) the vitreous. PC2, which accounts for 10% of the variance, implies an asymmetrical seesaw-like shape deformation of the ppBM layer and PC3 (5%) quantified the relative diameter of the BMO. 

Figure 3

 

Principal component analysis of shape variables from three groups of patients: (A) papilledema (open red circles, solid red circles), (B) normals (open squares), and (C) anterior ischemic optic neuropathy (+ symbols). This plot displays the first two principal components, which account for 88% of the variance in shape. The shape implied along PC1 (78%), displayed on the abscissa, is an anterior–posterior deformation that ranges in shape between an inverted-U shape (on the negative side) to a V-shape (on the positive). Along the ordinate, PC2 (10%) implies an alternating seesaw tilt around the BMO. Along PC1, patients with papilledema (solid and open red circles) tend to cluster on the negative side of the x-axis (within the red oval line) relative to normals (within the dotted circle) and AION (within black oval line), which tend to cluster on the right side. This demonstrates that the shape in papilledema is displaced anteriorly compared to normals and AION, which tend to be oriented posteriorly. Along PC2, papilledema patients in abduction (solid red circles) tend to cluster on the negative side of the y-axis and in adduction (open red circles) cluster on the positive side of the y-axis. This difference along PC2 shows that in abduction there is an anterior displacement temporally and posterior displacement nasally. The pattern reverses in adduction. The difference in eye position, for normals and AION, is not color coded in this Figure but illustrated in Figure 4. PC, principal component; n, nasal; t, temporal; V, v-shape; ∩, inverted-U-shape; ∼, alternating seesaw tilt shape.

Figure 4

 

This plot displays the PC2 scores (y-axis) for each group, in abduction and adduction, displayed along the x-axis. The shapes implied along PC2 are displayed in the left column as curved black lines relative to the mean shape (black circles). The PC scores for each patient are shown in the main body of the plot on the right. The paired scores in abduction (red circle) and adduction (black triangle) for each subject are linked by a red or black vertical line. If the deformation changed in a positive direction when going from abduction to adduction, then the difference is shown with a vertical red line. Changes in a negative direction are shown with a vertical black line. This Figure shows that for all groups, with few exceptions, adduction is associated with a relative posterior displacement temporally and a relative anterior displacement nasally. The shape pattern of differences reverses in abduction. The Figure also shows that differences in shape in papilledema is substantially greater than normals and AION.

The key differences between and within each cohort can be shown with the first two principal components (Fig. 3). Patients with papilledema tend to cluster on the negative side of the abscissa, whereas normals and AION tend to cluster on the positive side. Although there is some overlap among all three groups, the shapes of the ppBM layer among patients with papilledema are distinctly segregated from the other two with respect to PC1. This finding shows that the ppBM layer (presumably the eye wall itself) in patients with papilledema is anteriorly displaced or flattened compared to normals and AION, who generally have a V-shape oriented away from the vitreous. The difference with respect to PC1 by ANOVA between papilledema and the other two groups was statistically significant (P < 0.001). 

The axial raster images for each of the patients with papilledema are shown in Figure 3 so that the solid red circles represent the shape in abduction and the open red circles in adduction. Notwithstanding some overlap between the two groups, there also is a clear difference in shape of the ppBM layer based on eye position. On adduction, patients with papilledema tend to cluster in the lower left quadrant of the plot and on abduction in the upper left quadrant; that is, they segregate along PC2. More specifically, patients with papilledema in abduction show a relative anterior displacement temporally and posterior displacement nasally; on adduction this pattern is reversed. 

In contrast to papilledema, the overlap in shapes between abduction and adduction among normals and AION was considerable and, thus, abduction and adduction are not shown in Figure 3. However, a detailed paired comparison of each subject along PC2 is plotted in Figure 4, which shows each subject pair (in abduction and adduction) by group. In papilledema, there is a large and consistent positive shift along PC2 (shown with red lines) between abduction and adduction for all but 2 of the subjects. This shift reflects this seesaw shape difference between abduction (solid red circles) and adduction (black triangles). Normals and patients with AION showed a similar shift along PC2, although the magnitude of this shift was smaller. The mean difference in Procrustes distance units (a measure of the magnitude of shape difference) between abduction and adduction along PC2 for papilledema was 0.0435 ± 0.0308 (95% confidence interval [CI], 0.034–0.053), for normals 0.0066 ± 0.0055 (95% CI, 0.0040–0.0092), and for AION 0.0145 ± 0.0153 (95% CI, 0.0059–0.0229). The difference between papilledema and the other two was statistically significant (ANOVA, P < 0.001); the difference between AION and normals was not statistically significant. 

Individual examples of the change in shape induced by horizontal eye movements are demonstrated in Figures 5 through 7 with their respective descriptions. Figure 5, displayed in a magnified scale (3:2 aspect ratio), shows a typical example from each of the groups of patients. These show that the changes in papilledema are much greater than the small changes seen in AION and normals. Figure 6, displayed at a true scale (9:2, no magnification), shows the intermediate deformation that occurs between abduction and adduction by including primary position. The illustration in the sidebar provides a possible mechanism by which the CSF compartment may induce these changes, explained more fully in the discussion below. Figure 7 is an example of a patient showing that when papilledema recedes and intracranial pressure is normalized, the gaze induced deformation resolves. It is noteworthy that in all cases of papilledema shown in these Figures the temporal side is displaced posteriorly on adduction and the nasal side is displaced anteriorly, whereas in abduction the temporal side is displaced anteriorly and the nasal side moves posteriorly. These individual examples are consistent with the mean shapes in Figure 2. 

Figure 5

 

Spectral-domain OCT 9 mm axial rasters showing the difference between abduction and adduction in three patients: papilledema (a, b), ischemic optic neuropathy (e, f), and a normal subject (c, d). Please note that this Figure is displayed in a 3:2 aspect ratio and, thus, the vertical scale is magnified to demonstrate small changes in normals and AION. Bottom inset shows the corresponding superimposed tracings of the ppBM layer in abduction (red line tracing) and adduction (black line tracing). Although the direction of the difference is the same among all groups, the magnitude is greater in the patient with papilledema than the other two.

Figure 6

 

Spectral-domain OCT, 9 mm axial rasters, right eye, displayed in a true aspect ratio (9:2) from a 32-year-old woman with idiopathic intracranial hypertension and papilledema in abduction (top), primary position (mid), and adduction (bottom). The left column illustrates how gaze-induced shifts in CSF might affect the shape of the eye wall. The bottom inset shows superimposed tracings of the ppBM layer from each of the eye positions. In primary position there is a slight baseline asymmetry between the shape of the temporal and nasal BM layer that increases in adduction with a relative posterior displacement temporally and a relative anterior displacement nasally resulting in an S-shape configuration. The pattern is reversed in abduction.

Figure 7

 

Spectral-domain OCT, 9 mm axial rasters, right eye, displayed in a true aspect ratio (9:2) from a 30-year-old woman with idiopathic intracranial hypertension and papilledema in abduction (top row) and adduction (second row), at baseline before treatment (left column) and 6 months later after treatment with Diamox and resolution of the papilledema (right column). Bottom row shows superimposed tracings of the ppBM layer (red line is abduction, black is adduction). At baseline, in adduction there is a relative posterior displacement of the temporal BM layer and anterior displacement nasally. The pattern of displacement reverses in abduction. Six months later, with resolution of papilledema and normalization of the intracranial pressure, the overall anterior displacement and the gaze-induced deformations have resolved.

Discussion

This study reconfirms previous reports showing that the shape of the ppBM layer in patients with intracranial hypertension is displaced anteriorly (toward the vitreous), whereas normals and AION have a V-shaped ppBM layer oriented away from the vitreous. This particular characteristic is a consequence of the CSF pressure because the shape of the ppBM layer will shift posteriorly when CSF pressure is reduced (e.g., after lumbar puncture, shunting, or medical treatment) and anteriorly when the CSF pressure increases.^14,15 The present study extended these observations by demonstrating that horizontal eye movements can alter the shape of the globe in papilledema by causing a relative “seesaw” tilting deformation of the ppBM layer. On adduction, there is a relative posterior displacement (away from the vitreous) of the ppBM layer temporal to the BMO and an anterior displacement nasally. Overall, the deformation is S-shaped. On abduction, the shape pattern is reversed. The difference in shape between abduction and adduction in papilledema was statistically significant (permutation, P = 0.001). 

Normal subjects and patients with AION also display a difference in shape between abduction and adduction, but smaller in magnitude. Where shape difference in papilledema deforms both sides of the BMO, most of the differences in shape among normals and AION appears to predominantly affect the temporal side. Paired comparisons between abduction and adduction were significantly different at P = 0.03 (permutation) for normals and P = 0.01 (permutation) for AION. 

The mechanical effects of intracranial pressure on the eye principally target the optic nerve head and posterior pole of the globe. Intracranial hypertension increases the retrolaminar tissue pressure,^26,27 which obstructs axoplasmic flow,^28,29 expands the volume of the optic nerve head and radially displaces the juxtapapillary retina.^30–32 It also imposes a ring load on the scleral flange that distends the optic nerve sheath³³ and anteriorly displaces the peripapillary eye wall.^14,15 When these forces exceed a critical threshold, they also may cause folds in the retina and choroid that extend beyond the optic nerve head into the macula itself.^34,35 Subretinal and intraretinal fluid also may distort the microarchitecture of the retina.³⁰ 

The gaze-evoked shape differences in papilledema were greater than the shape changes in AION, and yet there was no significant difference in the degree of optic disc edema between the two. Moreover, there was no difference in shape deformations between AION (with disc edema) and normals. Thus, optic disc swelling alone does not explain the gaze-evoked disparities in shape between papilledema and AION. It is likely that these deformations in papilledema are the result of increased CSF pressure on the globe. Horizontal gaze changes the angle between the optic nerve, sheath, and globe. This redistributes the CSF ring load on the scleral flange by shifting fluid nasally in adduction and temporally in abduction (see Fig. 6, left side). An example of how the presumed fluid shift and the associated change in shape decreases when intracranial pressure and papilledema resolve is shown in Figure 7. 

In addition to the presumed effects of CSF pressure on the scleral flange, the material properties of the optic nerve sheath itself also may have a role. For example, patients with infiltrating tumors of the optic nerve sheath displace the ppBM layer^24,36 anteriorly presumably because the optic nerve sheath is stiffened. In mechanical systems, liquid-filled cylinders under high pressure are stiffer and more resistant to buckling^37,38 than low pressure or hollow cylinders. Thus, increased CSF pressure in the perioptic subarachnoid compartment could, in theory, cause a hydraulic stiffening (and distension) of the optic nerve sheath, which, in adduction, may tether the globe temporally and compress the globe nasally. The shape pattern reverses in abduction. It is likely that the CSF pressure load and stiffening of the optic nerve sheath have a major role among other factors. 

The clinical importance of gaze-induced deformations on the optic nerve has not been fully defined to our knowledge; however, there are two reports of gaze-evoked amaurosis in patients with intracranial hypertension and papilledema. The episodes were precipitated by horizontal gaze in either direction and resolved with medical treatment in one case and optic nerve sheath fenestration in the other.^39,40 Gaze-evoked amaurosis in papilledema, like orbital tumors, is caused by compression of the short postciliary artery branches and the circle of Zinn-Haller within the peripapillary sclera and or conduction block.^41,42 

The deformations in AION and normals demonstrates the effects of eye movement on the optic nerve and loadbearing structures without the confounding effects of elevated CSF fluid pressure. The component of shape deformation that might be influenced by the presence of disc edema is reflected in the difference between AION and normals. Although there appeared to be a greater shape difference in AION than in normals, the difference between the two groups was not statistically significant (P = 0.07, t-test). The number of patients in both groups is too small to draw any definite conclusions. 

Sigal et al.⁴³ showed that the principal determinant of laminar deformation and cupping in glaucoma is the peripapillary sclera. It is likely that these gaze-evoked peripapillary deformations are transmitted to the lamina cribrosa. The optic nerve also is strained along its length. Finite element analysis predicts that gaze-evoked strains on the optic nerve are considerable and may be influenced by the stiffness of the optic nerve sheath.⁴⁴ Anatomic and magnetic resonance imaging (MRI) studies have shown that in adduction the optic nerve is straight, whereas in abduction and primary position the nerve appears slightly bent and lax.^4,10,12 Friedman⁴ estimated that on adduction the temporal margin of the optic nerve is stretched 1.25 mm and with 50° abduction it is stretched 0.75 mm. More recently, using MRI, Demer¹² has shown straightening of the optic nerve develops at 26° adduction beyond which the globe is tethered and retracted by the optic nerve sheath rather than the optic nerve itself. Additionally, at the extremes of ocular duction the optic nerve is kinked at the scleral canal, which induces a tensile strain on the outer side of the bend and compression on the inner side. As has been suggested by Purkinje, and other investigators,^{2,4–7,12,44} it appears that adduction may be mechanically more stressful than abduction, which may explain why the difference in shape among normals and AION consists of a greater posterior displacement temporally than nasally. 

The difference in shape between abduction and adduction, in normal subjects and AION, was small but “statistically significant”; which is not to suggest that this is “clinically significant.” The “customary ocular motor range” is a product of ocular and head rotation⁴⁵ and so the extremes of ocular rotation are rarely used (perhaps with the exception of some patients with high myopia or large angle tropias). Moreover, traction on the optic nerve itself with ocular rotations may be shielded by the tensile limits of the optic nerve sheath in adduction.¹² It also is known that the viscoelastic response of the sclera is time-dependent, becoming stiffer with rapid spikes in stress, shielding the optic nerve head from large sudden deformations that might be elicited with large saccades.^46,47 

Conversely, static deformations of the globe may not reflect the dynamic effects of eye movements on the optic nerve, loadbearing structures, or blood supply. For example, numerical modeling shows that saccadic eye movements may generate multiple deformational waves that propagate posteriorly to the BMO, where they are reflected and summate with incoming waves after which they rapidly dissipate.⁴⁸ 

There is psychophysical evidence to support this idea of transient gaze-evoked strains on the optic nerve and juxtapapillary retina. The “fiery rings” described by Purkinje in 1823, expanded on by Helmholtz in 1911² and others,^3,5–7 are momentary, pericecal phosphenes elicited by adducting saccades or convergence. They appear as crescents on the temporal margin of the blind spot and are best seen in the morning in the dark. The “flick phosphene” described by Nebel⁵ is similar and gaze-evoked phosphenes also have been reported in patients with optic neuritis.⁴⁹ Both Purkinje and Helmholtz² deduced that “fiery rings” originate from a sudden stress on the optic nerve. Friedman^4,6 suggests that on adduction, the peripapillary retina is mechanically deformed by tensile forces temporally and compressive forces⁵⁰ nasally. Nebel⁵ speculated that these gaze-evoked phosphenes were due to vitreopapillary traction. However, Enoch et al.⁷ have shown that the “fiery rings can be evoked in subjects with a complete posterior vitreous detachment. Demer¹² suggests that on adduction much, if not all, of the strain takes place temporally predominantly borne by the optic nerve sheath that protects the optic nerve from tensile loading.⁵¹ 

That adduction phosphenes are located commonly on the temporal side of the blind spot (i.e., the nasal side of the optic disc), that they may occasionally encircle the blind spot, and also may be evoked on abduction suggests that dynamic strains probably are not limited to one side of the optic nerve head or one direction of gaze. The characteristics of gaze-evoked phosphenes described above also suggests that the dynamic effects of eye movements may be greater than the static deformations might otherwise indicate. 

Whether these gaze-evoked deformations are functionally significant is an open question. However, there are several potential pathways by which horizontal eye movements might functionally impact the optic nerve and juxtapapillary retina. Repetitive motion injuries have been implicated in a variety of neural and orthopedic disorders.⁵⁰ The degree of peripapillary scleral strain may affect blood flow though the branches of the short posterior ciliary arteries as they pass through sclera to supply the lamina cribrosa and prelaminar axons.⁵¹ Lastly, transmembrane integrin receptors, located in the lamina cribrosa,⁵² bind to the extracellular matrix and respond to mechanical stimuli by signaling intracellular cytoskeletal structures that can modulate cell structure and function.^53–57 

The findings in this study, and others cited, raise questions about the potential role of eye movements in the genesis or progression of a number of optic nerve and retinal disorders. In addition to papilledema and AION discussed in this paper, biomechanical factors also may be important in optic disc drusen, peripapillary subretinal hemorrhages, neovascular membranes, myopic degeneration, pigment epithelial detachments in optic nerve pits, the complications of ocular hypotony, and glaucoma to name but a few. 

Acknowledgements

The author thanks Mark J Kupersmith, MD and Randy Kardon, MD for their critical review and comments; Mary Mladek and Renee Jones for the expert SD-OCTs and fundus photography; and Ann Marie Lavorna for coordinating the IRB processes. 

Disclosure: P.A. Sibony None