QUT Digital Repository:
Shaw, Alyra J. and Collins, Michael J. and Davis, Brett A. and Carney, Leo G. (2009)
Eyelid pressure: inferences from corneal topographic changes. Cornea, 28(2).
© Copyright 2009 Lippincott, W illiams & W ilkins
Eyelid pressure: inferences from corneal topography changes
Alyra J. Shaw BAppSc(Optom), Michael J. Collins PhD, Brett A. Davis BAppSc, Leo G.
Contact Lens and Visual Optics Laboratory, School of Optometry,
Queensland University of Technology,
Victoria Park Road, Kelvin Grove, Queensland, Australia, 4059
Phone: +61 7 3138 5715, Fax: +61 7 3138 5665, E-mail: email@example.com
Financial interest of the authors: none
Figures: 5 Table: 1
Date of initial submission: 29/01/2008
Date of revised submission: 08/07/2008
Date accepted: 11/08/2008
Purpose: It is known that eyelid pressure can influence the corneal surface. However the
distribution of eyelid pressure, the eyelid contact area and the biomechanics of the changes
are unknown. While these factors are difficult to directly measure, analysis of eyelid-induced
corneal topography changes and eyelid morphometry enable some inferences to be drawn.
Methods: Eighteen subjects, aged between 19 and 29 years, with normal ocular health
were recruited. Corneal topography changes were measured after four conditions consisting
of two downward gaze angles (20° and 40°) and two types of visual tasks (reading and
steady fixation). Digital photography recorded the width of Marx’s line, the assumed region
of primary eyelid contact with the cornea. Results: Significantly larger corneal changes
were found after the 40° downward gaze conditions compared with 20° due to the upper
eyelid contact (p<0.001). For the 40° downward gaze tasks, the lower eyelid changes were
greater than those due to the upper eyelid (p<0.01). The upper eyelid Marx’s line width was
associated with the amplitude of corneal change (R²=0.32, p<0.05). Conclusion: Analysis
of the corneal topography changes gives insight into the pressure applied by the upper and
lower eyelids in different situations. These include greater upper eyelid pressure with
increasing downward gaze and larger lower eyelid pressure compared to the upper eyelid in
40° downward gaze. There was some evidence that supports Marx’s line as the primary site
of contact between the eyelid margins and the cornea.
Key words: corneal topography, eyelid pressure, near tasks, Marx’s line
The cornea is the principal refractive component of the eye. However, the corneal surface is
not fixed and is known to be susceptible to eyelid pressure. Increased or altered eyelid
pressure can occur from eyelids that are abnormal in structure. For example, induced
astigmatism from eyelid chalazia and haemangiomas has been shown to be dependent on
the size and location of the eyelid defect1, 2. Ptosis also causes changes in corneal
astigmatism 3, 4, with surgery able to partially reverse the induced change 4-7.
Normal eyelids can also alter corneal topography. Eyelid retraction and narrowing of the
palpebral aperture have been shown to alter corneal astigmatism 8-10. Typically the
mechanical effect of the eyelids on the cornea causes a change in 90/180 astigmatism 8, 11.
In primary gaze the vertical palpebral aperture is 9.7 mm, which decreases to 7.9 mm in 20°
downward gaze and 6.4 mm in 40° downward gaze 12. The altered position of normal eyelids
on the cornea in downward gaze has been found to have a clear association with bands of
corneal topography change which are parallel to the eyelid margin13-15.
Marx’s line extends along both the upper and lower eyelid margins and is thought to be the
primary site of contact between the eyelid margin and the surfaces of the bulbar conjunctiva
and cornea. It is a distinct line of squamous cells located just posterior to the Meibomian
gland orifices and is visible when stained with lissamine green dye 16, 17. Its anatomical
features are consistent with tissue that is subject to mechanical trauma and is thought to
stain with vital clinical dyes due to a lack of mucous coating. Doughty et al. 17 reported that
Marx’s line seems to be present in nearly all individuals and is an anatomical feature with no
obvious age or gender-related differences.
Previous investigations considering eyelid-induced corneal changes have assessed the
influence of downward gaze angle18, eye movements18 and different visual tasks including
reading, microscopy and computer work 14. An association between the magnitude of the
corneal optical changes and anatomical features of the eyelid has also been examined. Both
Buehren et al. 19 and Shaw et al. 20 found that narrower palpebral apertures during reading
lead to increased higher order aberrations and increased spherical power changes
respectively. Shaw et al. 20 have shown that eyelid curvature and tilt also influence the
magnitude of corneal optical change. It was also noted that subjects with narrower palpebral
apertures during reading did not necessarily have significantly narrow aperture sizes in
primary gaze 19.
This study provides further detail of the corneal changes induced by the upper and lower
eyelids. A localised topographical analysis method was developed to determine the
magnitude, shape and width of the corneal changes. Analyzing the eyelid-induced corneal
changes for a number of conditions allowed the relative eyelid pressure to be inferred.
Digital photographs were taken of Marx’s line to explore possible correlations with the
MATERIALS AND METHODS
Eighteen subjects, aged between 19 and 29 years, with an average age of 23 ± 3 years took
part in the study. There were eight emmetropic (± 0.25 D) and ten myopic (≤ -0.50 D)
subjects and equal numbers of males and females. The average refractive error was -1.22 ±
0.66 D and ranged from +0.25 to -5.75 D with -0.75 D or less astigmatism. Inclusion criteria
for subjects included corrected acuity of 0.0 log MAR or better and normal anterior ocular
health. Conditions that excluded subject participation were blepharitis, conjunctivitis,
Meibomian gland dysfunction and dry eye, as they have the potential to alter eyelid
morphometry, ocular surface health and the staining of Marx’s line 21-23. A diagnosis of dry
eye was made on the basis of dry eye symptoms (McMonnies score ≥ 14) and reduced tear
film stability (NITBUT < 10 seconds) 24.
Subjects were excluded if they had a history of ocular surgery or rigid gas permeable contact
lens wear, to minimise existing topographical alterations from eyelid pressure and contact
lenses. All subjects were asked to avoid any substantial reading on the morning of the
experiment 14, 25 and soft contact lenses wearers were requested to refrain from lens wear for
at least 32 hours prior to testing 26-31. Only the left eye was tested for each subject to avoid
issues associated with enantiomorphism in eyelid morphometry and corneal topography 32, 33.
The protocol fulfilled the requirements of the university ethics committee and adhered to the
tenets of the Declaration of Helsinki, including obtaining informed consent from all
participants. Six baseline corneal topography measurements of the left eye were taken using
the Medmont E300 Corneal Topographer (Medmont Pty. Ltd., Victoria, Australia). This
videokeratoscope has high levels of accuracy and precision for both spherical and aspherical
test surfaces 34 and high repeatability on human subjects 35. Subjects were instructed to
blink prior to topography measurements and the image was taken within a few seconds.
The protocol was designed to allow investigation of the effects of eyelid pressure in four
downward gaze conditions. This involved combinations of two downward gaze angles, 20°
and 40°, with two visual tasks, reading (eye movements) and steady fixation (no eye
movements). The 20° downward gaze conditions most closely represents the vertical eye
gaze angle usually adopted during reading (approximately 25°) 36, while the 40° conditions
examine a more extreme downward gaze angle. Blinking occurred during both the reading
and steady fixation conditions. While blink rate has been shown to vary for different tasks 37-
39, variation in blink rate is unlikely to influence the corneal changes. A blink lasts
approximately 0.26 seconds 40 and therefore with approximately 11.1 blinks per minute 41,
less than 5% of the total time is spent blinking. So the four conditions studied were: 1)
reading at 20° downward gaze; 2) steady fixation at 20° downward gaze; 3) reading at 40°
downward gaze; and 4) steady fixation at 40° downward gaze.
For each of these tasks the subject was positioned in a head rest, to ensure consistency of
eye and head position, and viewed a computer monitor with reading text at 40 centimetres
distance from the eyes. For the 15-minute reading task, three lines of N12 text (standard
novel font size) were visible at any one time through a cut-out window and a mouse or
keyboard was used to scroll the text to allow continuous reading. This set-up ensured that
the subject’s eye movements were primarily in the horizontal plane. During the steady
fixation tasks a fixation cross was visible in the centre of the cut-out window. Subjects were
instructed to blink naturally during all four tasks, which were completed in random order.
Immediately following the near task, six topography measurements were captured. As the
magnitude of lid-induced corneal distortions declines sharply immediately after a near task 42,
these measurements were taken within two minutes.
A lissamine green impregnated strip, wetted with two drops of sterile saline, was applied to
the superior and inferior bulbar conjunctiva of the left eye. To photograph Marx’s line the
upper eyelid was everted and the lower eyelid was rotated away from the eye globe. Images
of Marx’s line were captured using a high resolution digital camera and macro lens with a
fluorescent ring light.
Following data collection, height data were exported from the videokeratoscope and
analysed using custom written software. Elevation topography maps were chosen for
analysis as these most closely represent the actual physical corneal change. An average of
6 maps were exported and averaged according to the method of Buehren et al. for each
condition 25, Elevation difference maps (post-task minus pre-task) were then calculated to
examine eyelid-induced corneal changes after each visual task.
Within the band of topography change, the corneal valley (depression) was used as the
reference. User-defined points were chosen along the valley and fit with a 4th order
polynomial function (Figure 1). Cross-sections of data were taken perpendicular to the
polynomial function, extending 1 mm either side, with a spacing of 0.05 mm. The limits of
these cross-sections define the localised area under consideration (refer to open circles and
first and last cross-sections in Figure 1). Across each 2 mm cross-section, 16 points were
interpolated with a spacing of 0.13 mm, replicating the Medmont topographer’s normal radial
spacing of data points. Each cross-section was fit with a 9th order polynomial, which was the
lowest order found to minimise the fit error while also closely representing the original data.
The maxima and minima (peaks and valleys) were determined by locating when the first
derivative of the fit was equal to zero.
As a measure of the magnitude of the eyelid-induced corneal change, peak-to-valley
amplitudes of change were calculated from the difference topography maps 42. Peak-to-
valley amplitudes were determined from the valley towards the centre of the cornea (central
peak-to-valley) and from the valley towards the edge of the cornea (peripheral peak-to-
valley) (Figure 2). The distance from the central peak to the peripheral peak was used as a
measure of the width of corneal change. The amplitudes and widths for each subject and
condition were saved in an output file along with the original x, y and z coordinates of the
valley and the corresponding cross-section number. Three corneal regions were defined
relative to the videokeratoscope centre: nasal (-2.5 to -1.5 mm), central (-0.5 to 0.5 mm) and
temporal (1.5 to 2.5 mm) (Figure 1). Approximately 20 cross-sections within each of these
areas were averaged to obtain mean values for each region.
Some subjects had narrow palpebral apertures and couldn’t open their eyes wide enough, so
the eyelash shadows often inferred with the image captured by the corneal topographer. For
these subjects the eyelid-induced changes were often at the edge of the topography map.
Data was only analysed for subjects with distinct eyelid-induced corneal changes that were
not in close proximity to the edge of the topography map. This ensured the integrity of the
polynomial fit to the elevation topography data. The number of maps analysed for the upper
eyelid-induced corneal change for each condition was 9 (20° reading), 10 (20° steady
fixation), 13 (40° reading) and 17 (40° steady fixation). For the lower eyelid there were no
eyelid-induced corneal changes within the area captured by the videokeratoscope for the 20°
tasks. While for both the 40° reading and 40° steady fixation conditions, 11 maps had
complete eyelid-induced changes and were analysed.
The central and peripheral peak-to-valley amplitudes were compared using mixed linear
analysis and their relationship examined by a Pearson’s product moment correlation. The
ideal method to analyse the effect of the downward gaze angle, type of visual task, corneal
region and central peak-to-valley amplitude versus peripheral peak-to-valley amplitude,
would be to apply a repeated measures MANOVA. However there was a high correlation
between central and peripheral peak-to-valley amplitudes and few subjects had data for
every condition and corneal region. So a mixed linear model was used which took into
account the differing amount of data per subject by including subject identity as a random
factor. Using this method, the effects of downward gaze angle, visual task and corneal
region were investigated for the central peak-to-valley amplitudes for both upper and lower
eyelids. A comparison of upper and lower eyelid-induced corneal changes was examined
using Pearson’s correlation. Associations between the amplitude of corneal change and the
corneal topography simK values (mean K, flat K, steep K and astigmatism) were also
investigated with Pearson’s correlations.
Mixed linear analysis was also applied to the width of corneal change (central peak to
peripheral peak width) for both the upper and lower eyelid-induced corneal changes.
Correlations were investigated between the peak-to-peak width of the 20° and 40° tasks and
between the widths after the reading and steady fixation tasks.
Custom-written software was used to measure the width of Marx’s line of the upper and
lower eyelids in the nasal, central and temporal regions using the digital photographs (Figure
3). Five measurements of width were made in each of the three regions and the averages
calculated for both the upper and lower eyelids. For both eyelids the average standard
deviation between repeated measures of Marx’s line width was approximately 20% of the
mean (about 0.02 mm). There were small variations in the width of Marx’s line between
nasal, central and temporal locations. However the error associated with estimating Marx’s
line width was of similar magnitude to the regional variations, so only the central Marx’s line
width was used in subsequent analyses. Associations between the central Marx’s line width
and the corneal topography peak-to-peak widths and peak-to-valley amplitudes were
investigated with Pearson’s correlations.
Peak-to-valley amplitudes of corneal change
The mean peak-to-valley amplitudes were calculated for all conditions, corneal regions and
for both eyelids. In the central corneal region the mean upper eyelid-induced central peak-
to-valley amplitudes were 1.4 ± 0.6 µm (20° reading, n=9), 1.4 ± 0.5 µm (20° steady fixation,
n=10), 1.9 ± 0.7 µm (40° reading, n=13) and 1.5 ± 0.5 µm (40° steady fixation, n=17) (Figure
4). The corresponding peripheral peak-to-valley amplitudes were larger than those for the
central cornea for each condition: 1.6 ± 0.8 µm (n=5), 1.6 ± 0.3 µm (n=5), 2.5 ± 1.4 µm (n=9)
and 1.7 ± 0.8 µm (n=14) respectively (Figure 4). For the lower eyelid there were no corneal
changes within the area captured by the videokeratoscope for the 20° tasks. For the 40°
reading and 40° steady fixation tasks, the mean lower eyelid-induced central peak-to-valley
amplitudes were 2.8 ± 1.5 µm (n=11) and 2.6 ± 1.1 µm (n=11) and the peripheral peak-to-
valley amplitudes were 2.2 ± 1.4 µm (n=2) and 2.5 ± 0.8 µm (n=4) respectively (Figure 5).
For the upper eyelid region there was a highly significant difference between the central and
peripheral peak-to-valley amplitudes (mixed linear analysis, p<0.001), with the mean
peripheral peak-to-valley amplitudes being larger than the central peak-to-valley amplitudes
(Figure 4). For the lower eyelid-induced corneal changes there was no statistical difference
between the central and peripheral peak-to-valley amplitudes (Figure 5). The central and
peripheral peak-to-valley amplitudes were highly correlated for both the upper and lower
eyelid-induced changes (R²=0.69, p<0.001). Due to this strong correlation, further analyses
were conducted only using the central peak-to-valley amplitudes.
All the within-subject factors (downward gaze angle, type of task and corneal region) were
found to have a significant influence on the central peak-to-valley amplitudes for the upper
eyelid. The corneal changes after 40° downward gaze were larger than after 20°, changes
after reading were larger than those following steady fixation and nasal corneal changes
were larger than the central and temporal regional changes (Figure 4, Table 1). However
this was not the case for the region affected by the lower eyelid. Analysis of the 40° tasks
showed that neither the type of visual task nor the corneal region (nasal, central or temporal)
were statistically significant factors in the elevation change (Figure 5, Table 1). As there was
a consistent amplitude of change across the cornea (nasal, central and temporal regions),
only data from the central region data is presented in Figures 4 and 5.
For the 40° tasks, there were statistically significant differences in peak-to-valley amplitudes
between the upper and lower eyelids, with the lower eyelid causing greater elevation change
(p<0.01). There was also some weak evidence of an association between the magnitude of
corneal changes induced by the upper and lower eyelids for an individual, with a positive
correlation coefficient (R2=0.27, p<0.1). There were no significant associations between
corneal curvature (mean K, flat K, steep K or astigmatism) and the magnitude of the eyelid-
induced corneal change for any of the task conditions (p > 0.05).
Peak-to-peak width of corneal change
The mean distances from the central peak to the peripheral peaks due to the upper eyelid
were 1.3 ± 0.2 mm (for the 20° reading, n=5 and 20° steady fixation conditions, n=5) and 1.4
± 0.2 mm (for the 40° reading, n=9 and 40° steady fixation conditions, n=14). The lower
eyelid peak-to-peak widths were 1.4 ± 0.3 mm (n=2) and 1.2 ± 0.1 mm (n=4) for the 40°
reading and 40° steady fixation tasks respectively. The mixed linear analysis showed that
the downward gaze angle was the only significant factor (p<0.01) affecting the upper eyelid-
induced peak-to-peak width, with the width of change after the 20° tasks being smaller than
after the 40° downward gaze conditions. There were no statistically significant differences
related to the type of visual task or the corneal region (nasal, central or temporal) analysed
for either upper or lower eyelid. The upper eyelid peak-to-peak width for the 20° steady
fixation task was significantly correlated with the peak-to-peak width for the 40° steady
fixation task (R²=0.60, p<0.01). While there were no significant correlations between the
widths after the 20° and 40° reading tasks or between reading and steady fixation conditions.
Marx’s line width
The average width of Marx’s line centrally for the upper eyelid was 0.11 ± 0.05 mm and 0.13
± 0.10 mm for the lower eyelid. There were no statistically significant correlations between
the widths of the upper and lower eyelid central Marx’s line, or between the Marx’s line width
and the peak-to-peak width of corneal change for either eyelid for each individual. There
was however, some evidence of an association between the upper eyelid Marx’s line width
and the peak-to-valley amplitudes. A positive correlation for the 40° steady fixation task
(R²=0.32, p<0.05) indicated that a wider Marx’s line was associated with a deeper corneal
change. There was also a positive correlation for the 20° steady fixation condition peak-to-
valley amplitudes and Marx’s line width (R² = 0.14) though it did not reach statistical
significance. So there was evidence of an association between Marx’s line and eyelid-
induced corneal changes.
We have used a localised topographical analysis approach to investigate eyelid-induced
corneal changes after downward gaze tasks. As previously reported, the presence of the
eyelids on the corneal surface in downward gaze causes wave-like corneal distortions 13, 14, 18,
25, 42, 43. The average elevation central peak-to-valley amplitudes ranged from 1.4 to 2.8 µm.
As longer reading periods have been shown to cause larger corneal changes 42, this result
for 15-minute tasks is comparable to the 4 µm average recorded for one hour reading
The magnitude of the downward gaze angle has a significant impact on the induced corneal
change. For the upper eyelid, elevation peak-to-valley amplitudes were 25% larger for the
40° conditions compared with those at 20°. Centrally, the upper eyelid was 0.6 mm closer to
the videokeratoscope centre for the 40° tasks than the 20° tasks. It is doubtful whether the
corneal biomechanical properties would alter enough over this small distance to make the
corneal tissue more susceptible to lid pressure and account for the 25% increase in corneal
change. It is more likely that at the increased downward angle, the upper eyelid exerts
higher pressure on the ocular surface. This increased pressure may be due to the eyelid
resting closer to the centre of the cornea on a region of ‘higher cornea’ nearer to the apex.
As Collins et al. have previously discussed, this would only be valid if there was no change in
the antero-posterior position of the eye with downward gaze 18. It is well known that the eye
globe retracts on blinking with the contraction of the orbicularis oculi 40, 44. In contrast, a
downward gaze lid saccade is achieved almost exclusively by the passive elastic forces of
the ocular tissues and the relaxation of the levator palpebrae superioris muscle with no
involvement of the orbicularis oculi 45. As these eyelid movements have different
mechanisms it is unlikely that the globe would retract with a downward eyelid saccade.
There is a possibility that the extraocular muscles may influence the corneal shape at large
downward gaze angles. However this seems unlikely, as previous studies have failed to
conclusively show a change in corneal shape with convergence 46-48. Therefore it seems
that eyelid pressure on the cornea is greater when the lid is closer to the corneal centre in
larger downward gaze angles.
In contrast, despite the lower eyelid changes being further from the videokeratoscope centre
(2.7 mm compared with 1.5 mm for the upper eyelid), the corneal changes associated with
the lower eyelid were deeper than those associated with the upper eyelid. The lower eyelid
has previously been shown to produce larger corneal changes than the upper eyelid
following a 45° downward gaze task 18, although the position of the eyelids on the cornea
was not considered. These findings suggest that at large downward gaze angles the lower
eyelid pressure on the corneal surface is greater than that of the upper eyelid.
Previous studies have only been concerned with the central peak-to-valley amplitude of the
wave-like change 14, however in this study both central and peripheral peak-to-valley
amplitudes were analysed. There was a high correlation between the central and peripheral
peak-to-valley amplitudes. For the upper eyelid, corneal changes were greater peripherally
compared with centrally. There is little known about the exact area of contact between the
eyelid margins and the cornea. The upper eyelid margin has been described as a `lid-wiper’
21, aiding tear film distribution during blinking. Mathematical modelling of the upper eyelid’s
blinking action suggests that the margin needs to change angle during opening and closing
to effectively distribute tear fluid 49. The shape or angle of the upper eyelid margin as it
contacts the cornea could result in unequal tangential forces and asymmetrical tissue
distribution. In contrast, the lower eyelid induced symmetrical peak-to-valley profiles with
equal peaks either side of the valley and its movement during blinking is primarily in the
horizontal direction 40.
For the lower eyelid, movements associated with reading did not produce statistically
significant differences in corneal changes compared to the steady fixation conditions.
However for the more mobile upper eyelid, reading-associated movements increased the
14 Download full-text
peak-to-valley amplitudes by 3% and 25%, for the 20° and 40° downward gaze tasks
respectively. Similar increases of between 17% and 35% have been recorded for 1 Hz eye
movements compared with a steady fixation control for 15 minute visual tasks 18. The
increased amplitude of change may be due to increased frictional force between the upper
eyelid and the ocular surface. There was evidence that eye movements during reading
produced more random spatial corneal changes since the upper eyelid peak-to-peak widths
of the 20° and 40° steady fixation conditions were correlated, while there were no significant
correlations between the widths for the reading tasks. This conclusion was also supported
by the larger standard deviations of corneal change associated with the reading conditions
compared with the steady fixation tasks.
While the pressure of the eyelids is a major factor in the induced corneal change, corneal
factors may also be involved. There was a moderate positive correlation between the upper
and lower eyelid peak-to-valley amplitudes at 40° downward gaze. Subjects with a large
upper eyelid-induced change were more likely to have a large lower eyelid-induced change.
This indicates that certain individuals may be more susceptible to the pressure from the
upper and lower eyelids due to either the cornea’s mechanical properties or from common
anatomical features of the eyelids.
Due to its anatomical structure, it is thought that Marx’s line is the natural site of frictional
contact between the eyelid margin and the surfaces of the bulbar conjunctiva and cornea.
The average width of the upper eyelid Marx’s line in this group of subjects was 0.11 ± 0.05
mm, which is a similar to the previous report of 0.10 ± 0.09 mm 16. There does not appear to
be any previous published values reporting the width of the lower eyelid Marx’s line. The
measured width of 0.13 ± 0.10 mm in this study suggests that it has similar dimensions to the
While there was no association evident between Marx’s line width and the width of the