·Clinical Research·
Comparisons
of ganglion cell-inner plexiform layer loss patterns and its diagnostic
performance between normal tension glaucoma and primary open angle glaucoma: a
detailed, severity-based study
Xiao-Yu
Xu, Kun-Bei Lai, Hui Xiao, Yi-Quan Lin, Xin-Xing Guo, Xing Liu
State Key Laboratory of
Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University, Guangzhou,
Guangdong Province, China
Correspondence to: Xing Liu. State Key Laboratory of
Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University, 7 Jinsui
Road, Tianhe District, Guangzhou 510623, Guangdong Province, China.
drliuxing@163.com
Received:
Abstract
AIM: To evaluate the
patterns of macular ganglion cell-inner plexiform layer (GCIPL) loss in normal
tension glaucoma (NTG) and primary open angle glaucoma (POAG) in a detailed,
disease severity-matched way; and to assess the diagnostic capabilities of
GCIPL thickness parameters in discriminating NTG or POAG from normal subjects.
METHODS: A total of 157 eyes of
157 subjects, including 57 normal eyes, 51 eyes with POAG and 49 eyes with NTG
were enrolled and strictly matched in age, refraction, and disease severity
between POAG and NTG groups. The average, minimum, superotemporal, superior,
superonasal, inferonasal, inferior, and inferotemporal GCIPL thickness, and the
average, superior, temporal, inferior, and nasal retinal nerve fiber layer
(RNFL) thickness were obtained by Cirrus optical coherence tomography (OCT).
The diagnostic capabilities of OCT parameters were assessed by area under
receiver operating characteristic (AUROC) curves.
RESULTS: Among all the OCT
thickness parameters, no statistical significant difference between NTG group
and POAG group was found (all P>0.05). In discriminating NTG or POAG
from normal subjects, the average and inferior RNFL thickness, and the minimum
GCIPL thickness had better diagnostic capabilities. There was no significant
difference in AUROC curve between the best GCIPL thickness parameter (minimum
GCIPL) and the best RNFL thickness parameter in discriminating NTG (inferior
RNFL; P=0.076) and indiscriminating POAG (average RNFL; P=0.913)
from normal eyes.
CONCLUSION: Localized GCIPL loss,
especially in the inferior and inferotemporal sectors, is more common in NTG
than in POAG. Among all the GCIPL thickness parameters, the minimum GCIPL
thickness has the best diagnostic performance in differentiating NTG or POAG
from normal subjects, which is comparable to that of the average and inferior
RNFL thickness.
KEYWORDS: normal tension
glaucoma; primary open angle glaucoma; spectral domain optical coherence
tomography; ganglion cell-inner plexiform layer thickness; pattern
DOI:10.18240/ijo.2020.01.11
Citation:
Xu XY, Lai KB, Xiao H, Lin YQ, Guo XX, Liu X. Comparisons of ganglion
cell-inner plexiform layer loss patterns and its diagnostic performance between
normal tension glaucoma and primary open angle glaucoma: a detailed,
severity-based study. Int J Ophthalmol 2020;13(1):71-78
INTRODUCTION
Glaucoma is an optic neuropathy with
progressive loss of retinal ganglion cells (RGCs) and their axons that lead to
peripapillary retinal nerve fiber layer (RNFL) loss and glaucomatous visual
field damage[1]. Of all the glaucoma cases
worldwide, approximately 74% are primary open angle glaucoma (POAG)[2-3]. Although rise in intraocular
pressure (IOP) is regarded as the primary risk factor of glaucoma progression[4], RGC loss and glaucomatous optic neuropathy can occur
partially independently of IOP in normal tension glaucoma (NTG), usually known
as a subset of POAG.
Conventionally, an IOP of less than
Spectral
domain optical coherence tomography (SD-OCT) enables measurements of both
peripapillary RNFL and macular thickness parameters, which has been widely
accepted as a standard of care in managing glaucoma. The enhanced scanning
speed, better image resolution, and improved retinal layer segmentation ability
of the ganglion cell algorithm (GCA; Cirrus Version 6.0; Carl Zeiss Meditec,
Dublin, CA, USA) enable the measurement of the macular ganglion cell-inner
plexiform layer (GCIPL) thickness. The glaucomatous diagnostic ability of
macular GCIPL thickness has been confirmed to be comparable to or better than
that of the peripapillary RNFL thickness by multiple studies[10-14]. One of the explanations is that the GCIPL thickness
is expected to target at the RGCs, which are primarily affected by glaucoma,
directly in an area of their highest concentration[15-18]. However, there is a paucity of studies on the
differences in the distribution and the discriminating capability of macular
OCT measurements between NTG and POAG. The aims of this study were to compare
macular GCIPL parameters measured by Cirrus OCT between age-, refraction-, and
severity-matched NTG and POAG, and to investigate and compare the diagnostic
performance of GCIPL and RNFL thickness parameters in differentiating NTG and
POAG from normal eyes.
SUBJECTS AND METHODS
Ethical Approval All study subjects were consecutively
recruited at Zhongshan Ophthalmic Center of Sun Yat-sen University, Guangzhou,
China from August 2014 to May 2015. The study was approved by the Institutional
Review Board (IRB) and followed the tenets of the Declaration of Helsinki.
Written informed consent was obtained from each subject.
Study Subjects All subjects were performed complete
ophthalmic examinations including visual acuity (uncorrected and
best-corrected), refraction examination (cycloplegic refraction test was
performed if the participant was <30 years of age, whereas manifest refraction
test was performed if the participant was ≥30 years of age), central corneal
thickness (CCT) measurement (Ultrasonic Pachymetry DGH-1000, Storz Inc, Louis,
MO, USA), slit lamp biomicroscopy, angle evaluation using gonioscopy, fundus
examination, fundus photography (Kowa nonmyd a-D III; Kowa Optimed Inc, Aichi,
Japan), Humphrey perimetry (SITA standard 24-2; Carl Zeiss Meditec, Dublin, CA,
USA), and Cirrus HD-OCT (Carl Zeiss Meditec, Dublin, CA, USA). IOP was measured
by the same well-trained investigator using the calibrated Goldman applanation
tonometer (Haag-Streit, Bern, Switzerland) in the day time (
Glaucoma diagnosis was made if
characteristic structural changes to the optic disc and RNFL defects
accompanied by glaucomatous visual field defects were found. Glaucomatous
visual field defect was defined as: glaucoma hemifield test outside the normal
range, pattern standard deviation with P <5%, or a cluster>3
points in the pattern deviation plot in a single hemifield (superior or
inferior) with P<5%, and at least one of these should have P<1%.
A reliable visual field testing should have a false-positive error, a
false-negative error, and a fixation loss of less than 20%, simultaneously.
Eyes with these features were diagnosed as POAG when IOP>
The inclusion criteria of this study
included: age≥18y, refractive diopters (spherical equivalence) between -3 to +3
D, 360-degree open angle, reliable visual field results, and OCT scans with
good quality. The exclusion criteria were unsuccessful image acquisition,
history of ocular trauma, usage of medications that could cause elevation of
IOP or optic neuropathy, and any life-threatening diseases. For eyes with
glaucoma, they also included: any known history of ocular disorders other than
age-related cataract, diseases that might affect retina health and visual field
results, and history of intraocular surgery. If both eyes of a subject met the
criteria, only one randomly selected eye was enrolled.
Optical Coherence Tomography
Scanning Eyes were dilated with 0.5%
tropicamide and 0.5% phenylephrine before OCT scans. OCT scans of macular cube
512×128 protocol and optic disc cube 200×200 protocol were performed with the
same Cirrus OCT device by a well-trained ophthalmologist (Xu XY). Images with
signal strength of less than 7 or those with visible motion or blinking
artifacts and segmentation failure were considered of poor quality and
discarded immediately.
The average, minimum, and sectoral
(superior, superonasal, inferonasal, inferior, inferotemporal, and
superotemporal) macular GCIPL thickness was calculated using the GCA algorithm
within a
Statistical Analysis SPSS version 22.0 (SPSS Inc.,
Chicago, IL, USA) was used for statistical analysis. Kolmogorov-Smirov test and
Levene test were conducted to test the normality and homogeneity of variance,
respectively. Chi-square test was used to evaluate the differences of gender
distribution among groups. Age, refractive diopters, values of mean deviation,
CCT, IOP, GCIPL and RNFL thickness were compared with one-way analysis of
variance (ANOVA), with Bonferroni adjustments used for further pairwise
multiple comparisons. Independent t-test was used for comparisons
between the anti-glaucoma drops being used, OCT parameters and visual field
mean deviation between early NTG and early POAG groups, and moderate NTG and
moderate POAG groups. The diagnostic ability of each OCT parameter was
determined by the area under the receiver operating characteristic (AUROC)
curve, which were then compared using MedCalc 18.0 (Med-Calc Statistical
Software bvba, Mariakerke, Belgium) based on the method of DeLong et al[19]. A P value of <0.05 was considered to be
significant statistically.
RESULTS
A total of 157 eyes were included.
The baseline demographic and clinical characteristics of subjects in all 3
groups were showed in Table 1. There were no significant differences in age,
gender, refractive diopters or CCT among NTG, POAG, and normal groups.
Base-line IOP was significantly different among the 3 groups, while the
difference between NTG and normal group was insignificant in pairwise
comparisons (P=0.767). Although the value of MD was significantly
different among the 3 groups, no significant difference was found between NTG
and POAG groups in pairwise comparisons (P=0.812).
Table 1 Demographic and clinical
characteristics of the study population
mean±SD
Parameters |
NTG (n=49) |
POAG (n=51) |
Normal (n=57) |
P |
|
Age (y) |
62.2±13.2 |
60.7±12.1 |
61.9±12.5 |
0.819 |
|
Gender (male/female) |
19/30 |
28/23 |
29/28 |
0.244 |
|
Spherical equivalent (diopter) |
-1.87±1.81 |
-1.81±1.37 |
-1.58±1.66 |
0.118 |
|
CCT (μm) |
547.8±30.7 |
558.2±33.6 |
554.1±27.1 |
0.440 |
|
Baseline IOP |
13.10±3.80 |
24.93±4.12 |
14.30±3.51 |
<0.001 |
|
Anti-glaucoma drops |
1.33±0.47 |
1.55±0.40 |
- |
0.041 |
|
Disease stage (early/moderate) |
28/21 |
29/22 |
- |
0.977 |
|
MD (dB) |
|
|
|
|
|
Total |
-6.04±3.77 |
-6.20±3.96 |
-0.87±0.82 |
<0.001 |
|
Early stage |
-3.26±1.65 |
-3.30±1.77 |
- |
0.933 |
|
Moderate stage |
-9.75±2.27 |
-10.04±2.48 |
- |
0.692 |
|
NTG: Normal tension glaucoma; POAG:
Pimary open angle glaucoma; CCT: Central corneal thickness; IOP: Intraocular pressure;
MD: Mean deviation. Comparisons among 3 groups (NTG, POAG and normal control
groups) were performed using one-way ANOVA. Comparisons between NTG and POAG
groups were performed using independent t-test.
The GCIPL and RNFL thickness of each
group were displayed in Table 2, with the NTG and POAG groups further
subdivided into 3 categories of total (patients of all severity stages), early
stage and moderate stage. Though the differences in all GCIPL and RNFL
thickness and ONH parameters were of statistical significance among 3 groups
except that of disc area, no statistically significant difference between the
NTG and the POAG group was found in all parameters using pairwise comparisons
with Bonferroni adjustments. Pairwise comparisons of the differences of all
parameters between NTG and normal, and POAG and normal showed significant
differences (all P<0.001) except that of disc area (between NTG and
normal, P=0.064; between POAG and normal, P>0.999). OCT
parameters between early NTG and POAG, and moderate NTG and POAG were also
compared and displayed in Table 2.
Table 2 Macular GCIPL and
peripapillary RNFL thickness in all study groups
mean±SD
parameters |
NTG |
POAG |
Normal (n=57) |
P values |
|||||||
Total (n=49) |
Early (n=28) |
Moderate (n=21) |
Total (n=51) |
Early (n=29) |
Moderate (n=22) |
Pa |
Pb |
Pc |
Pd |
||
GCIPL thickness (μm) |
|||||||||||
Average |
69.78±10.04 |
72.71±9.61 |
65.48±9.13 |
70.43±8.06 |
73.72±7.14 |
65.33±7.37 |
83.96±5.43 |
<0.001 |
>0.999 |
0.653 |
0.928 |
Minimum |
58.16±13.72 |
62.75±13.18 |
52.13±11.66 |
60.78±10.28 |
64.24±10.11 |
55.50±8.73 |
80.84±5.86 |
<0.001 |
>0.999 |
0.633 |
0.207 |
Superotemporal |
70.49±10.71 |
72.54±10.75 |
66.70±10.43 |
70.12±11.26 |
74.52±9.86 |
63.33±10.59 |
82.58±5.45 |
<0.001 |
0.612 |
0.471 |
0.283 |
Superior |
72.18±13.03 |
74.50±11.46 |
68.57±14.09 |
72.67±10.35 |
75.66±9.81 |
67.67±10.14 |
84.79±6.14 |
<0.001 |
>0.999 |
0.684 |
0.924 |
Superonasal |
75.88±11.81 |
78.00±9.84 |
72.52±13.27 |
76.22±10.05 |
77.52±9.77 |
73.75±10.23 |
86.33±5.98 |
<0.001 |
>0.999 |
0.853 |
0.699 |
Inferonasal |
71.04±13.48 |
74.07±12.44 |
66.61±13.49 |
70.35±9.21 |
73.69±8.16 |
65.33±8.67 |
83.88±6.14 |
<0.001 |
>0.999 |
0.891 |
0.773 |
Inferior |
64.65±12.51 |
68.75±11.34 |
59.09±11.59 |
66.33±9.75 |
70.07±9.02 |
60.71±8.63 |
82.26±5.67 |
<0.001 |
>0.999 |
0.628 |
0.494 |
Inferotemporal |
64.37±11.26 |
68.11±11.29 |
59.65±8.95 |
66.33±10.47 |
70.66±9.81 |
60.13±8.35 |
83.68±6.26 |
<0.001 |
0.900 |
0.367 |
0.648 |
Peripapillary RNFL thickness (μm) |
|||||||||||
Average |
74.61±9.91 |
78.29±6.52 |
68.26±12.05 |
73.22±11.90 |
77.79±10.54 |
66.38±11.13 |
96.93±8.67 |
<0.001 |
>0.999 |
0.833 |
0.468 |
Superior |
92.22±21.73 |
96.75±17.34 |
83.78±25.79 |
91.14±19.03 |
97.86±16.81 |
80.71±18.43 |
119.60±14.62 |
<0.001 |
>0.999 |
0.807 |
0.571 |
Temporal |
61.67±11.52 |
63.96±12.03 |
57.13±11.11 |
56.96±10.82 |
58.34±10.02 |
53.58±12.43 |
69.68±9.03 |
<0.001 |
0.076 |
0.060 |
0.307 |
Inferior |
81.27±17.10 |
88.57±13.54 |
70.70±16.18 |
84.10±23.49 |
93.34±21.97 |
71.83±19.83 |
127.98±17.15 |
<0.001 |
>0.999 |
0.330 |
0.945 |
Nasal |
63.20±9.42 |
63.93±8.59 |
61.13±10.77 |
60.20±8.49 |
61.28±7.55 |
58.83±9.87 |
70.26±9.73 |
<0.001 |
0.318 |
0.220 |
0.267 |
Optic nerve head parameters |
|||||||||||
Rim area (mm2) |
0.893±0.147 |
0.939±0.122 |
0.807±0.174 |
0.910±0.258 |
0.997±0.192 |
0.795±0.281 |
1.367±0.252 |
<0.001 |
>0.999 |
0.178 |
0.618 |
Disc area (mm2) |
2.202±0.542 |
2.195±0.481 |
2.225±0.606 |
2.056±0.432 |
2.067±0.434 |
2.058±0.424 |
1.992±0.416 |
0.064 |
0.351 |
0.296 |
0.311 |
Vertical cup-to-disc ratio |
0.738±0.086 |
0.707±0.096 |
0.788±0.052 |
0.710±0.127 |
0.676±0.118 |
0.765±0.127 |
0.462±0.185 |
<0.001 |
0.964 |
0.285 |
0.395 |
GCIPL: Ganglion cell-inner plexiform
layer; RNFL: Retinal nerve fiber layer; NTG: Normal tension glaucoma; POAG:
Primary open angle glaucoma. aComparisons were performed using
one-way ANOVA among NTG, POAG and normal groups. bComparisons of NTG
and POAG were performed using Bonferroni adjustments based on one-way ANOVA. cComparisons
were performed using independent t-test between early NTG and early POAG
groups. dComparisons were performed using independent t-test
between moderate NTG and moderate POAG groups.
The differences of the 6 sectoral
GCIPL thickness parameters and the average GCIPL thickness were calculated and
compared between NTG and POAG groups, whose result as well as the proportions
of the average, minimum, and 6 sectoral GCIPL thickness of NTG group and POAG
group to normal group were showed in Table 3.
Table 3 Patterns of GCIPL loss in
NTG and POAG
mean±SD
parameters |
Differences between each sectoral and the average
GCIPL thickness |
Thickness percentage (glaucomatous/normal eyes) |
|||
NTG (μm) |
POAG (μm) |
P |
NTG (%) |
POAG (%) |
|
Average |
- |
- |
- |
83.10 |
83.88 |
Minimum |
- |
- |
- |
71.95 |
75.19 |
Superotemporal |
0.71±6.67 |
-0.31±6.02 |
0.420 |
85.36 |
84.91 |
Superior |
2.41±6.50 |
2.23±6.16 |
0.892 |
85.13 |
85.70 |
Superonasal |
6.10±6.03 |
5.78±7.00 |
0.809 |
87.89 |
88.28 |
Inferonasal |
1.27±6.46 |
-0.08±4.74 |
0.237 |
84.70 |
83.88 |
Inferior |
-5.12±6.33 |
-4.10±6.37 |
0.422 |
78.59 |
80.64 |
Inferotemporal |
-5.41±9.27 |
-4.10±6.54 |
0.418 |
76.92 |
79.27 |
GCIPL: Ganglion cell-inner plexiform
layer; NTG: Normal tension glaucoma; POAG: Primary open angle glaucoma.
Table 4 showed the AUROC curves with 95% confidence interval
(CI) of the OCT parameters and MD values of visual field testing. In
discriminating NTG from normal eyes, the minimum GCIPL thickness and the
inferior RNFL thickness were the parameters with the best diagnostic capability
in all GCIPL and RNFL parameters, respectively. There was no significant
difference of AUROC curve between the minimum GCIPL thickness and the inferior
RNFL thickness (Z=-1.776, P=0.076). For diagnosing eyes with POAG
from normal eyes, the best parameters were the minimum GCIPL thickness and the
average RNFL thickness, respectively. Also, no significant difference in AUROC
curve was found between these two parameters (Z=0.109, P=0.913).
In each OCT parameter as well as the MD value, there was no statistically
significant difference in AUROC curve between NTG and POAG.
Table 4 Diagnostic capabilities of
OCT parameters and visual field MD for discriminating NTG or POAG from normal
subjects
Parameters |
NTG |
POAG |
P value |
||
AUROC |
95%CI |
AUROC |
95%CI |
||
GCIPL thickness |
|
||||
Average |
0.917 |
0.857-0.977 |
0.933 |
0.879-0.988 |
0.697 |
Minimum |
0.950 |
0.908-0.993 |
0.960 |
0.924-0.997 |
0.731 |
Superotemporal |
0.874 |
0.797-0.950 |
0.866 |
0.788-0.943 |
0.886 |
Superior |
0.856 |
0.774-0.938 |
0.856 |
0.776-0.936 |
1.000 |
Superonasal |
0.838 |
0.747-0.929 |
0.847 |
0.759-0.936 |
0.889 |
Inferonasal |
0.835 |
0.747-0.923 |
0.905 |
0.838-0.973 |
0.220 |
Inferior |
0.916 |
0.853-0.979 |
0.939 |
0.883-0.996 |
0.594 |
Inferotemporal |
0.924 |
0.866-0.981 |
0.918 |
0.856-0.980 |
0.890 |
RNFL thickness |
|
||||
Average |
0.982 |
0.961-1.000 |
0.957 |
0.917-0.996 |
0.273 |
Superior |
0.867 |
0.789-0.945 |
0.886 |
0.816-0.956 |
0.724 |
Temporal |
0.731 |
0.617-0.844 |
0.828 |
0.738-0.917 |
0.191 |
Inferior |
0.991 |
0.978-1.000 |
0.948 |
0.902-0.994 |
0.086 |
Nasal |
0.626 |
0.500-0.751 |
0.728 |
0.617-0.838 |
0.230 |
Mean deviation |
0.944 |
0.891-0.998 |
0.925 |
0.864-0.987 |
0.644 |
NTG: Normal tension glaucoma; POAG: Primary
open angle glaucoma; GCIPL: Ganglion cell-inner plexiform layer; RNFL: Retinal
nerve fiber layer; AUROC: Area under the receiver operating characteristic; CI:
Confidence interval.
DISCUSSION
In this study, the GCIPL and RNFL
thickness parameters between age-, refraction-, and severity-matched NTG and
POAG were evaluated. No significant difference was found in all OCT parameters
but the temporal RNFL thickness. The diagnostic performance of the GCIPL and
RNFL thickness parameters in discriminating eyes with NTG or POAG from normal
eyes were also investigated and compared. In discriminating NTG and POAG from
normal eyes, the minimum GCIPL thickness and the average RNFL thickness were
the best parameters among all GCIPL and RNFL parameters, respectively, and they
also demonstrated comparable diagnostic performance in discriminating NTG and
POAG.
It is widely accepted that NTG is a
subtype of the spectrum of POAG and it accounts for a significant percentage of
open-angle glaucoma[20-21].
NTG was defined as progressive optic neuropathy and glaucomatous visual field
defects, with an untreated maximum IOP of
While the IOP-independent causative
factors are commonly associated with the development of NTG, the risk factors
are not yet completed verified. Older age, being females, thinner CCT, myopia,
genetic background, and systemic vascular diseases including migraine, low
blood pressure, low diastolic ocular perfusion pressure, Alzheimer disease,
primary vascular dysregulation, Flammer syndrome, metabolic syndrome,
obstructive sleep apnea syndrome and others are known risk factors of NTG[33-35]. Previously published
literatures suggested that eyes with NTG tend to have greater RNFL thinning
inferiorly and inferotemporally than superiorly and more preserved temporal
RNFL quadrant[36]. Some studies showed RNFL thinning
was more localized in NTG patients compared to POAG patients[36-37]. Kim et al[38]
also found that ganglion cell complex (GCC; which is the sum of macular RNFL,
GCL, and IPL thickness) loss was more localized in NTG group and more diffuse
in POAG group. However, the relationship of the risk factors and the
characteristics of NTG including the susceptibility of the inferior RNFL fibers
and the localized thinning pattern of RNFL and GCC remained unclear.
In our study, the minimum GCIPL
thickness was thinner in NTG group than in POAG group, although not necessarily
be statistically significant, which indicated an obvious localized thinning of
GCIPL in eyes with NTG. It could be the inferior and/or inferotemporal sectoral
GCIPL that account for the major localized thinning effect among all sectoral
GCIPL locations, which was consistent to the previous findings showing the more
localized inferior or inferotemporal peripapillary RNFL defects in NTG than in
POAG[38]. The more obvious localized thinning of
GCIPL in NTG may indicate more diffuse thinning of GCIPL in POAG group,
especially in the superior hemisphere. In NTG group and POAG group with similar
visual field mean deviation, Kim et al[38]
and Jung et al[39] found that GCC or GCIPL
in the superior hemisphere was significantly thinner in POAG group, serving as
a compensation of the thinning in the inferior hemisphere in NTG group. Some
other studies found no significant thinning in certain locations between NTG
and POAG groups[40-41].
Although our study did not show significant sectoral GCIPL thinning, the
discretization in each sectoral GCIPL parameters, however, were greater than
that of POAG, supporting the facts of localized thinning that other
observations have showed.
The introduction of “minimum” GCIPL thickness
(any one of the 360 degrees) in our study may suggest that the pattern of GCIPL
thinning in NTG group does not necessarily mean that the localized thinning has
to occur in inferior and/or inferotemporal sector. The location where localized
GCIPL thinning occurs could be highly variable between cases. Therefore,
minimum GCIPL thickness may preserve more information about the localized
thinning, being more sensitive than those averaged sectoral GCIPL thickness
parameters (averaged thickness of 60 degrees) for eyes with NTG, which can
interpret why the difference of the minimum GCIPL thickness was more
significant than other sectoral GCIPL thickness between NTG and POAG group.
Despite this difference, other sectoral GCIPL thickness parameters were similar
between NTG and POAG, suggesting that these two subtypes of glaucoma have
similar distribution pattern of ganglion cells in the macular.
Using a strict age-, refraction-,
and disease severity-matching strategy, it was not surprising that the
diagnostic capabilities of OCT parameters didn’t show much difference in
discriminating NTG or POAG from normal subjects in this study. Like other
studies[10-14,39],
our results also suggested that the minimum GCIPL thickness, the average and
the inferior RNFL thickness were the parameters with the best glaucoma
diagnostic capabilities. The disability of the use of IOP makes diagnosing NTG
much more challenging than diagnosing high-tension glaucoma in the early stage,
leading to its potentially severe underdiagnosis rate. Although there are a
number of studies have investigated and compared the OCT parameters, such as
RNFL thickness, macular thickness, GCC thickness between NTG and POAG, to the
best of our knowledge, only one of these studies were specifically focusing on
the diagnostic ability of GCIPL thickness in discriminating NTG and POAG and
found that the diagnostic performance was comparable in differentiating these
two subtypes of glaucoma[39]. Given the fact that
we included only early and moderate stage glaucoma patients in this study, the
AUROC curves of all GCIPL thickness parameters were higher than 0.80, showing
that the macular GCIPL thickness could be considered as a reliable objective
parameter in diagnosing NTG. Among these, the minimum, the inferotemporal, the
average, and the inferior GCIPL thickness had AUROC curves higher than 0.90,
which was in consistent with the findings that the average and the inferior
peripapillary RNFL thickness had better ability in diagnosing glaucoma than
RNFL thickness in other locations[11].
It is important to introduce a multivariable
model combining valuable information for making diagnosis from all available
OCT parameters, visual field parameters, as well as blood flow/retina vessel
density information revealed by OCT angiography into clinical use. The
application of artificial intelligence (AI) algorithms after deep learning
could be ideal for glaucoma detection which demands a highly personalized data
combination and analysis strategy. Future explorations including the
improvement in understanding and extracting the existing knowledge and
optimization of data selection for AI analysis will be one of the key steps in
revolutions of glaucoma diagnosing methodology.
There were several limitations of
this study. First, this was a hospital-based retrospective study. All NTG patients
were treated with anti-glaucoma drugs shortly after the diagnosis was made.
Longer observation was not allowed, while probably being ethically beneficial,
may potentially result in selection bias because those might eventually develop
to high-tension glaucoma if left untreated could be included as NTG at this
point. Second, the relatively small sample size and the lack of golden standard
for diagnosing extreme early stage glaucoma may miss preperimetric glaucoma,
making the distribution of GCIPL thickness analysis not representative enough
for the whole cohort. Investigations with a larger cohort are expected in the
future. Last, the disease severity evaluation was based on the cutoff points of
the MD of visual field testing. Such classification criteria may be too rough
to make the NTG group and the POAG group compatible in a precise
severity-matched way.
In summary, localized GCIPL loss,
especially in the inferior and inferotemporal sectors was more commonly seen in
NTG patients than in POAG patients. While there was no significant difference
in macular GCIPL thickness between the NTG group and the POAG group that were
strictly matched in age, refraction, and disease severity, the minimum GCIPL
thickness, the average and inferior RNFL thickness were the OCT parameters with
better diagnostic capabilities. Future investigations on an integral, AI-based
glaucoma diagnostic platform are needed to optimize early glaucoma detection
and management.
ACKNOWLEDGEMENTS
Foundations: Supported by National Natural
Science Foundation of China (No.81800879); Natural Science Foundation of
Guangdong Province (No
Conflicts of Interest: Xu XY, None; Lai KB, None; Xiao
H, None; Lin YQ, None; Guo XX, None; Liu X, None.
REFERENCES