Evaluation of Retinal Nerve Fiber Layer and Inner Macular Layers in Primary Open-Angle Glaucoma with Spectral-Domain Optical Coherence Tomography

The aim of our research is to assess and compare the peripapillary retinal nerve fiber layer (pRNFL) thickness diagnostic capability with those of three macular parameters— macular RNFL (mRNFL) thickness, GCL+ (ganglion cell layer with inner plexiform layer thickness), and GCL++ (mRNFL and GCL+) in primary open-angle glaucoma patients with spectral-domain optical coherence tomography (SD-OCT). The 414 participants (483 eyes) aged 45–84 years in this prospective study were recruited from Eye Clinic at the University Hospital “Alexandrovska” (Sofia, Bulgaria). They were divided into 6 groups: controls, ocular hypertension, preperimetric glaucoma (PPG), and three groups of perimetric glaucoma stages—early, moderate, and advanced. OCT was performed using Topcon 3D OCT 2000 device, as eight parameters from two protocols (Circle and Glaucoma Analysis—Macula) were analyzed. The results showed that the RNFL highest diagnostic capability parameter is Total mRNFL (AUROC 0.879 in PPG and 0.929 in early glaucoma stage). The macular highest diagnostic accuracy parameter was found GCL++ without any significance from mRNFL diagnostic possibilities. The results of current research showed that mRNFL possesses high diagnostic accuracy in comparison analysis with other pRNFL and macular OCT parameters in early glaucomatous changes. Macular RNFL and its highest diagnostic possibilities could be successfully used as an individual diagnostic parameter separated from the whole ganglion cell complex in the early glaucoma changes.


Introduction
According to the American Academy of Ophthalmology the medical term glaucoma is used for a group of diseases that damage the optic nerve (ON) as distinctive type of optic neuropathy characterized by structural (cupping of optic nerve head-ONH, changes in connective tissue structural elements and number of the nerve fibers in ON) and functional changes (typical visual field defects). Increased intraocular pressure (IOP) is one of the most common risk factors associated with developing and progressing of the disease, but its presence or absence does not change the above-mentioned glaucoma definition [1].
Epidemiological studies found that the glaucoma at the end of twentieth century covered more than 60 million people around the world. Prognostic studies show an increasing trend of the number of affected patients, and in 2020 it is going to be approximately 80 million people, and in 2040 approximately 112 million [2][3][4]. Cataract and glaucoma are leading causes of blindness worldwide. Because of the reversibility of the vision after cataract extraction, the glaucoma remains the leading cause of irreversible blindness. The large number of glaucoma patients, irreversible vision loss and the impact on the life quality of the affected people are just part of the reasons for making glaucoma one of the diseases with big social influence.
Glaucoma is characterized by irreversible loss of ganglion cells, which axons form the ON. Ganglion cells are localized in three retinal layers-inner plexiform layer or IPL (their dendrites), ganglion cell layer or GCL (their bodies), and retinal nerve fiber layer or RNFL (their axons). Therefore exactly the above-mentioned layers are those, which glaucoma affects accompanying typical visual field defects [5]. Chronic and progressive loss of neuroretinal tissue is cardinal feature of glaucomatous optic neuropathy (GON) and criterion for diagnosis [6].

Anatomical aspects of the RNFL
All afferent pathways in the ON start from a layer of photoreceptors (cones and rods), which is located in the retina on area more than 1000 mm 2 . In ONH all fibers are concentrated on surface with approximately 2-3 mm 2 area [7,8]. From the body of each ganglion cell comes out a nerve fiber or axon, which moves toward the ONH. So that it can be called conglomerate consisted of all converging axons, which are as mentioned above part of the retina and form a layer-RNFL [9]. Nowhere else the ganglions' nerve fibers are not so much compact as they are in ONH, and this is what determines the importance of peripapillary RNFL thickness in diagnosis and follow up of patients with GON.
These are some characteristics of RNFL [9, 10]: 1. Papillomacular nerve fiber bundle-it starts from ganglion cells in the foveolar region. The nerve fibers from nasal foveolar area move straight toward the temporal border of ONH, and those from the temporal part make a slight arc around the nasal nerve fibers and then join to the straight bundle.

2.
Superior and inferior retinal arcades-they are created by later formed nerve fibers and ganglion cells originating temporal to the fovea. They arc around the macula and papillomacular bundle to enter the ONH. Temporal raphe or seam-it extends from the fovea to the temporal part of the retina and  consists of very few axons, because by rule the nerve fibers from the upper half of the  retina do not pass the horizontal meridian to the arcuate course of the nerve fibers from the  lower part of the retina, and vice versa. 4. An extremely large collection of nerve fibers in the superior and inferior quadrant of the ONH-namely these two regions are set to be more vulnerable for glaucomatous damages.

5.
Nasal nerve fibers-they move radially toward the ONH.

6.
Exact location of the nerve fibers in the ONH according to their position in the fundus-the more peripheral retinal location the more central ONH localization.
Basic features, used to make an assessment of RNFL images [9]: 1. Striations of RNFL-normally RNFL can be seen as striated bright and dark lines in the areas of superior and inferior temporal blood vessels in healthy eyes. If atrophy is presented (<50 μm RNFL thickness) the striations of the background disappear and bright lines cannot be seen because of the RNFL loss [11] (see Figure 1).

2.
Defects of the background brightness-they can be diffuse loss and localized defects (wedge-shaped and cleft-shaped). The width in the cleft-shaped defects is the same along the full length, however the width in wedge-shaped defects is different, peripherally they are wider and become narrower toward the ONH. This could be explained with convergent course of the nerve fibers. Diffuse defects have an impact over the complete RNFL thickness in the fundus and also their diagnosis is more difficult from localized defects.

3.
Visualization of the blood vessels-normally RNFL covers retinal blood vessels. That's why small and medium-sized blood vessels have unclear contours and look misty. When RNFL atrophy appears, then blood vessels can be seen clearly because of the less covering from the nerve fiber layer.
All nerve fibers are arranged in a specific way in the ONH not only in each and every human being but also in each and every of the human eyes. Equal quantity nerve fibers may look in a different way in the borders of ONHs with dissimilar disc area, depth of the lamina cribrosa, and height of the scleral canal [12]. Equal functional capacitate could be presented by different looking structures and vice versa-equal looking structures could have different functional activity [13][14][15].
The RNFL thickness depends on: age, ethnicity, number and thickness of the nerve fibers, quantity of the glia, quantity of the blood vessels, disk area of the ONH, axial length of the eye (Ax). The thickness of the measured RNFL depends also on: the stage of peripapillary atrophy/conus myopicus, vitreoretinal tractions. The excavation (cupping) depends on: disc area, number and thickness of the nerve fibers, quantity of the glia [14]. Normally in the course of time the RNFL thickness decreases with age normally with 4000-5000 axons per year [16][17][18][19] and this is approximately 2.0 μm/decade or 0.2% per year at mean thickness 100 μm [20]. The ON consists of 700,000-1.4 million nerve fibers and the RNFL thickness in healthy people has a wide variety of a norm. The usage of absolute values restricts the process of distinguishing healthy from glaucoma patients [20]. Therefore some authors talk about "modulation of RNFL thickness"-it shows the relative loss of nerve fibers as difference between the biggest and smallest measured value of RNFL thickness in a retinal region of interest [21].

RNFL and glaucoma
When assessing glaucomatous damages it is appropriate to measure the RNFL thickness, because thinning of this layer correlated directly with ganglion cells loss, which is the basic pathophysiological event [22]. Evaluation of the RNFL thickness is important for early glaucoma diagnosis before appearing of the clinical manifestations of the disease. It is proven that 40-50% of the nerve fibers are should be dropped out before developing of the visual field defects [23]. Clinical evaluation of the RNFL with red-free photography shows that thinning of the layer can be seen in 60% of the pictures 6 years before appearing of clinical manifestation of the defects in visual field [24]. These facts show that structural changes occur before the functional ones. Typical visual field defects in glaucoma are nasal step, arcuate scotoma, paracentral scotoma, generalized depression, and progressive worsening of the indices of the standard automated perimetry (SAP) [25].
Sometimes in glaucoma visual field defects can be seen without appearance of structural glaucomatous changes. It is possible also in equal RNFL loss to be obtained a different clinical finding according to initial RNFL thickness. This could be explained with the following: visual field defects appear after 40% loss of the nerve fibers. Each man is born with different quantity nerve fibers. If a person owns very thick for human population RNFL, the loss of 40% nerve fibers probably will not give any significant results in optical coherence tomography (OCT)-the line thickness will be in the middle of the green zone, the zone shows lack of disease. Then this individual is going to have functional defects with normal RNFL thickness. If another person is born with thin RNFL, the loss of 40% nerve Optic Nerve fibers will give significant results-OCT line thickness will be close to the yellow zone or in the zone. Then this individual is going to have functional defects with pathological thin RNFL [14,15].
In the early glaucoma stage it is considered that the affected ganglion cells decrease their functional processes before they die and this leads to decreasing of the visual functions without an obvious structural changes. This is the reason why a patient has functional manifestations of glaucoma in combination with normal RNFL thickness [14].
The most distant nerve fibers from ONH originate exactly from these farthest parts of the ganglion cells in the retina and they are located deeply in the RNFL. They pass closely to the scleral edge and most peripherally in the ON [26]. These nerve fibers that originate from the closest to the ONH parts of the retina are located superficially in the RNFL and pass centrally in the ON. It is thought that the nerve fibers, which are located superficially in the RNFL, are more vulnerable in glaucoma, and their damage is associated with an enlargement of the blind spot.
It is also believed that chronically increased IOP leads to compression of the circulation of the Elschnig's border tissue and its atrophy. Then lamina cribrosa starts posteriorization. It is considered therefore that it is a reason for stretching and rupturing of the nerve fibers which are closest to the scleral edge. Only nerve fibers in prelaminar region can drop out consequently, because they are separated and not in bundles. The affecting of the nerve fibers is from peripheral to central region [26,27]. Unordered affecting of the nerve fibers can be seen in acute angle closure glaucoma.

Retinal nerve fiber layer and inner macular layers evaluation
in primary open-angle glaucoma with spectral-domain optical coherence tomography

Purpose
The aim of our research is to assess and compare the peripapillary RNFL (pRNFL) thickness diagnostic capability with those of three macular parameters-macular RNFL (mRNFL) thickness, GCL+ (ganglion cell layer with inner plexiform layer thickness), and GCL++ (mRNFL and GCL+) in primary open-angle glaucoma patients with spectral-domain OCT (SD-OCT).

Material
All participants (healthy volunteers and patients) included in current clinical study were examined in the university eye clinic of Alexandrovska Hospital, Sofia, Bulgaria for total period of time-a year and 3 months. This is a prospective observational study of 414 participants (483 eyes) aged 45-84 years (mean 66.7 ± 8.7), male-132, and female-282. All patients were distributed into six groups:

Methods
All patients underwent full ophthalmological examination including: a complete case history for eye and general diseases; family history; refraction and best corrected visual acuity; slit-lamp examination; indirect fundus biomicroscopy; contact ultrasound pachymetry (OcuScan RxP -Alcon, Forth Worth, Texas, USA); Goldmann tonometry; indirect gonioscopy (Goldmann three-mirror gonioprism/Shaffer classification, 1960); SAP -SITA Standard 24-2, HFA II (Carl Zeiss Meditec, Dublin, CA, USA) with near correction if necessary. Only reliable perimetry results with total error rate (loss of fixation and false-positive and false-negative results) lower than 25%. The stage of POAG changes was determined using Hodapp-Parrish-Anderson classification.
Optical coherence tomography: All patients underwent SD-ОСТ of both eyes with dilated pupils by one examiner using Topcon 3D OCT 2000 (FA plus) (Topcon Corporation, Japan), software version -8.11.
Only OCT protocols with scan quality over 50%, no artifacts from eye or body movements, blinking, and lack of macular pathology (edema, drusen, holes) were included in the analysis.
Statistical methods: For statistically significant were considered the differences with P values <0.05. We used descriptive, dispersion and ROC-analysis to evaluate diagnostic accuracy, specificity, and sensitivity. A comparison was made between Ist group and IInd, Ist and IIIrd and so long. With comparison analysis we searched for statistical significant difference between some of the parameters' values in specificity and sensitivity.

Results
The descriptive statistics can be seen in Table 1, and mean values of the all RNFL parameters in Table 2. In Table 3 can be seen the ROC analysis and the diagnostic capabilities of the eight OCT parameters in each group. The RNFL parameter with highest diagnostic potential in the groups-PPG (AUROC = 0.879), early (AUROC = 0.929), moderate (AUROC = 0.989) and advanced glaucoma (AUROC = 1.000) is Total mRNFL followed by Inf mRNFL and Inf pRNFL. The RNFL parameters with lowest diagnostic potential in all glaucoma stages are Nas pRNFL, Temp RNFL. A single RNFL parameter (Total mRNFL) was measured with highest diagnostic accuracy for glaucoma in all of its stages -from PPG to advanced glaucoma. In the current investigation, it is shown for the first time the higher diagnostic ability of macular RNFL from those of peripapillary RNFL. In OH group, we found that Inf mRNFL has the highest diagnostic possibilities but without any clinical significance, because none of the RNFL parameters change significantly in patients with OH in comparison with control group.
AUROC values allowed us to create ROC curves in all groups as we included the five RNFL parameters with the best results (Figures 3-6).
After that we used comparison analysis to demonstrate if statistical significant difference exists between diagnostic possibilities of RNFL parameters in all groups. In Table 4 can be seen only two significant differences in AUROC values between Sup mRNFL (0.907) and Total mRNFL, and between Total mRNFL (0.929) and Inf pRNFL (0.867). Although, we found the highest diagnostic potential in all glaucoma groups for Total mRNFL, our comparison analysis showed that these possibilities are not statistical significant with exception of the above-mentioned two examples. For instance, in the stage of PPG, we were not able to find any statistical significant differences between the best five diagnostic RNFL parameters ( Table 4), so they have equal abilities to diagnose glaucoma patients in this particular stage. In the stage of early POAG, we found significant difference in diagnostic abilities between Inf pRNFL and Total mRNFL, so that it would be better if the clinician uses not the first five but the four best diagnostic parameters from Table 3.

Optic Nerve
We evaluated also sensitivity, specificity and cut-off values for the same RNFL parameters. In PPG group there are two parameters with highest and almost equal values of the sensitivity and specificity-Total mRNFL (sensitivity-0.83, specificity-0.77) and Inf mRNFL (sensitivity-0.82, specificity-0.79). These two parameters keep their high and close values of sensitivity also in the group of early perimetric glaucoma: Total mRNFL-0.93 и Inf mRNFL-0.90. The parameter with highest value of specificity in the same group is Total pRNFL-0.89, and after it are these parameters: Total mRNFL-0.81 and Inf mRNFL-0.79. In the PPG group with highest values is Total mRNFL (sensitivity-0.97 and specificity-0.95), and after it is Inf mRNFL (sensitivity-0.94 and specificity-0.85). It is observed very small differences in the values between investigated parameters, which decrease in advanced glaucoma group. With highest sensitivity (1.00) and specificity (1.00) in advanced glaucoma group is Total mRNFL, and after it is Inf pRNFL (1.00; 0.99) and Inf mRNFL (0.99, 0.99).
In Table 5 can be seen the AUROC values of the macular parameters-Total mRNFL, Total GCL+ (ganglion cell layer/GCL + inner plexiform layer/IPL) and Total GCL++ (GCL + IPL + mRNFL) from protocol Glaucoma Analysis-Macula (see Figure 3, left). Lowest diagnostic  accuracy for glaucoma in all investigated stages possesses the parameter-GCL+. The highest area under the curve has GCL++ (0.919, 0.932) in PPG group, Total mRNFL in the moderate glaucoma group (0.989), and the both parameters reach maximal possibilities for diagnosis in advanced glaucoma group (1.000). We also applied comparison analysis to find significance in diagnostic capabilities (AUROC values) between macular parameters. The results from this analysis could be seen in Table 6. Significance can be seen in the values between Total mRNFL and Total GCL+, and between Total GCL++ and Total GCL+. We did not find a difference between Total mRNFL and Total GCL++. This mean that the whole ganglion cell    Optic Nerve 118 layer (consists of three sub-layers), which is presented by GCL++ parameter, has an equivalent diagnostic potential of those of Total mRNFL, which presents only one of the macular sublayers (the most inner layer consists of the nerve fibers). The less accurate diagnostic potential from macular OCT parameters we found for GCL++. Therefore we exclude this parameter as an accurate in glaucoma diagnosis.

Discussion
In the current research we found that the Topcon OCT parameter-Total mRNFL has the highest diagnostic accuracy in the very early stage of glaucoma, in which only structural changes could be seen (PPG). It is important to know its diagnostic possibilities compared with those of other OCT parameters, because it allows the clinicians to precise the early diagnosis, appropriate treatment and the most important for the patients-prevention of the vision loss. The  results showed that this parameter also has the highest diagnostic possibilities in all perimetric glaucoma stages. These conclusions we made only after comparative analysis in diagnostic accuracy between all OCT parameters (peripapillary and macular) had been applied.
There are not many researches, which investigate macular RNFL as a separate parameter not as a part of whole ganglion cell complex. Not enough data was collected about characteristics, correlations and diagnostic possibilities of mRNFL.
In 2005 for the first time was created software algorithm for automated segmentation of retinal layers in Stratus OCT (OCT III). It helped authors differentiate four macular layers-macular nerve fiber layer (mNFL); inner retinal complex (IRC) consisting of ganglion cells, inner plexiform layer and inner nuclear layer; outer plexiform layer (OPL); outer retinal complex (ORC), consisting of outer nuclear layer, inner and outer photoreceptor segments. When the authors investigated diagnostic accuracy they found the highest values in mNFL+IRC (0.97), and lowest in OPL (0.56). Diagnostic accuracy of OPL and ORC was significantly lower from mNFL, IRC, mNFL+IRC and circumpapillary nerve fiber layer (cpNFL) (p ≤ 0.01). They found that AUROC values of IRC, mNFL+IRC and cpNFL were significantly higher from whole retinal thickness (p ≤ 0.049). It was not found significant differences between parameters with best diagnostic possibilities-mNFL, IRC, mNFL+IRC and cpNFL (p ≥ 0.15). The two parameters-ORC and OPL were found also to have almost permanent thickness in patients with glaucoma in comparison with healthy volunteers [28].
In the beginning of the era "OCT diagnostics in glaucoma" was found that the whole retinal thickness decreases. Later with the initiation of spectral domain OCT (SD-OCT) in the clinical practice inner macular layers (mRNFL, GCL, IPL) were called a complex (ganglion cell complex-GCC), which consists of the bodies, dendrites and axons of the ganglion cells [29]. A self-evident fact is that GCC has significantly higher possibilities for glaucoma diagnosis than the thickness of the whole retina.
Mwanza et al. investigated diagnostic accuracy of GCIPL (ganglion cell + inner plexiform layer), RNFL ONH parameters [30]. They found that GCIPL diagnostic possibilities are between 0.918 and 0.956, and there values are comparable with the best diagnostic parameters-RNFL (between 0.933 and 0.939) and ONH parameters (0.910 and 0.962) without statistically significant difference between them.
There are two conceptions of ganglion cell loss in glaucoma. In the first-the dendrites die before the bodies, and the most resistant part of the cell of glaucoma damage are their  Table 6. Comparison analysis between macular parameters' AUROC values.

Optic Nerve
axons. Therefore, it is reasonable to investigate the thickness of GCL + IPL separately from mRNFL. On the other hand IPL consists of the dendrites not only the ganglion cells but also the bipolar cells, and it is believed as more correctly to measure the thickness of mRNFL+GCL together.

Conclusion
Peripapillary RNFL is a proved glaucoma diagnostic parameter and also ganglion cell complex. Predominantly of the glaucoma comparisons in diagnostic accuracy are between pRNFL and GCC in different OCT devices.
The current research investigate a new SD-OCT macular parameter-mRNFL and its diagnostic possibilities for different stages of POAG. It proves that mRNFL could be used in every day clinical practice of the ophthalmologist as independent parameter with very high diagnostic possibilities for early stages of glaucoma when only structural changes are visible.
Now we are working on creating of staging system based on Total mRNFL values (cut-off values) in each glaucoma group. It could give possibilities for the ophthalmologists to use the values of this parameter in everyday clinical practice to make diagnosis and follow-up of the glaucoma patients. This grading system will be the only of the OCT structural systems created up to date. Total mRNFL has the potential to be one of the best OCT diagnostic parameters and we as researchers must find how to use it in the diagnosis of very early glaucoma changes.

Author details
Bilyana Mihaylova* and Galina Dimitrova *Address all correspondence to: b51@abv.bg Department of Ophthalmology, Medical University of Sofia, University Hospital 'Alexandrovska', Sofia, Bulgaria