Comparative Study of Macular Ganglion Cell–Inner Plexiform Layer and Peripapillary Retinal Nerve Fiber Layer Measurement: Structure–Function Analysis

Abstract

Purpose.: We explored and compared the relationships between the visual field (VF) sensitivities assessed by standard automated perimetry (SAP), and the ganglion cell–inner plexiform layer (GCIPL) and peripapillary retinal nerve fiber layer (pRNFL) thicknesses as measured by Cirrus high-definition optical coherence tomography (HD-OCT) in glaucomatous eyes.

Methods.: We enrolled 213 eyes of 213 patients with glaucoma. The thicknesses of the average/sectoral GCIPL and pRNFL were measured by Cirrus HD-OCT. The mean sensitivity (MS) of 24-2 SAP was recorded on decibel and 1/L scales. The topographic relationships between structure and function were investigated.

Results.: Statistically significant correlations between the corresponding VF sensitivity and the macular GCIPL thickness were found in all GCIPL sectors. Among six GCIPL sectors, the strongest association was observed between superonasal center MS and inferotemporal GCIPL thickness. In comparative analysis, the association between the central cluster MS and average GCIPL thickness was significantly stronger than that of temporal pRNFL thickness using the decibel scale (P < 0.001). The association between regional VF sensitivities, and the inferior hemifield and inferior GCIPL thicknesses were significantly stronger than those of the corresponding pRNFL thickness using the decibel scale (P = 0.001 and 0.007).

Conclusions.: The average and sectoral GCIPL thicknesses determined by Cirrus HD-OCT were associated significantly with global and regional VF sensitivity in patients with glaucoma. The macular GCIPL thickness values may provide more valuable information than temporal pRNFL thickness values for understanding the structure–function relationships of the macular region.

Introduction

A papillomacular bundle defect preceding a defect in other sectors of the peripapillary retinal nerve fiber layer (pRNFL) may occur in glaucoma patients with baseline IOP in the statistically normal range. ¹ Parafoveal visual field (VF) defects can occur during the early stages of glaucoma, although parafoveal VF loss generally has been considered to occur in association with more advanced field defects. ^2–4 A previous study using experimental primate glaucoma models has shown that the loss of retinal ganglion cells (RGCs) within the central 12° of the retina may be present even in eyes with mild glaucomatous damage. ⁵ Clinicians should ascribe great importance to VF defects near the fixation point in glaucoma patients, because a central VF defect can adversely affect driving performance and reading ability. ^6,7 Sensitive methods for the detection of structural changes in the papillomacular fiber bundle in glaucomatous eyes may be crucial for understanding the structure–function relationship of the central retina. 

The axons of the RGCs within the macular region are assumed to be located mainly in the temporal sectors of the optic nerve head (ONH). However, the diagnostic usefulness of the temporal pRNFL thickness in distinguishing between normal and glaucomatous eyes is inferior to that of the superior and inferior pRNFL thicknesses of the four quadrants, ^8,9 and the regional VF sensitivity is more weakly associated with the temporal pRNFL thickness than with the superior and inferior pRNFL thicknesses. ^10,11 It might have been due, at least in part, to under-representation of the macular region in the 24-2 strategy. However, the crowding hypothesis of Hood et al. ¹² provides a possible explanation for this, in that the temporal sectors of the ONH are less susceptible to glaucomatous damage, compared to the superior and inferior regions. In addition, the decrease in temporal pRNFL thickness might not reflect accurately a loss of RGC because the temporal pRNFL generally is thinner than the superior and inferior pRNFL. Furthermore, anatomic variations, such as peripapillary atrophy, can create larger measurement errors in the temporal pRNFL. Therefore, the assessment of papillomacular fiber loss based on the pRNFL is limited. 

Since its introduction, optical coherence tomography (OCT) technology has undergone several improvements, and more recent enhancements of segmentation algorithms have enabled more precise quantitative assessment of retinal layers, particularly the macular area. Kim et al. ¹³ reported that pRNFL thickness and macular ganglion cell complex (GCC) thickness measured by Fourier domain (FD)–OCT (RTVue-100 GCC scan; Optovue, Inc., Fremont, CA), had similar structure–function relationships with VF sensitivity. A more recent study by Na et al. ¹⁴ showed that GCC thickness has a significant structure–function association with macular VF, and the association was significantly greater than that of the macular pRNFL with macular VF in the superior central VF area. 

Nevertheless, there is a concern that the inclusion of the RNFL may reduce the diagnostic value of these measurements. In contrast with GCC thickness measurements, the Cirrus high definition (HD)–OCT ganglion cell/inner plexiform layer (GCIPL) algorithm successfully yields the combined thickness of the RGC layer and the inner plexiform layer without including the RNFL, and measures GCIPL thickness values with excellent intervisit reproducibility. ^15,16 While a GCC scan by FD-OCT can provide only two hemifield thickness measurements (superior and inferior GCC thicknesses), the GCIPL segmentation algorithm can provide six sectoral GCIPL thickness measurements (superotemporal, superior, superonasal, inferonasal, inferior, and inferotemporal GCIPL thicknesses). Therefore, the GCIPL thickness measured by Cirrus OCT allows a more precise topographic analysis of the central retina than the pRNFL thickness, although the GCIPL sectors do not correspond perfectly to the distribution of the RGC axons to the ONH. The purpose of this study was to explore and compare the relationship between global and regional VF sensitivities as assessed by standard automated perimetry (SAP), and the average/sectoral GCIPL thickness and pRNFL thickness values measured by Cirrus HD-OCT in glaucomatous eyes. 

Patients and Methods

In this cross-sectional study, 213 eyes of 213 patients with a diagnosis of perimetric glaucoma were enrolled retrospectively from the clinical database at the glaucoma clinic of Seoul St. Mary's Hospital, College of Medicine, The Catholic University of Korea, between April 2012 and May 2013. This study was conducted in accordance with the ethical standards stated in the Declaration of Helsinki and with the approval of the Institutional Review Board of Seoul St. Mary's Hospital, College of Medicine, The Catholic University of Korea, Seoul, Korea. 

Upon initial evaluation, each participant underwent a comprehensive ophthalmologic examination, including a review of medical and ocular histories, measurement of best-corrected visual acuity, slit-lamp biomicroscopy, Goldmann applanation tonometry, gonioscopic examination, dilated funduscopic examination, stereoscopic optic disc photography, red-free RNFL photography, achromatic automated perimetry using the 24–2 Swedish Interactive Threshold Algorithm (SITA) standard program (Humphrey Visual Field Analyzer; Carl Zeiss Meditec, Inc., Dublin, CA) and spectral domain (SD)–OCT scanning (Cirrus HD-OCT; Carl Zeiss Meditec, Inc.). Participants included in this study met the following criteria: best corrected visual acuity of 20/30 or better, with a spherical equivalent between −6.0 and +4.0 diopters (D) and a cylinder correction within ±3.0 D, the presence of a normal anterior chamber, and open-angle on slit-lamp and gonioscopic examinations, and reliable VF test results with false-positive errors of <15%, false-negative errors of <15%, and fixation loss of <20%. Subjects were excluded on the basis of any of the following criteria: a history of any retinal disease; a history of ocular trauma or surgery, including trabeculectomy or glaucoma drainage device implantation, with the exception of uncomplicated cataract surgery; other optic nerve disease except for glaucoma; and a history of systemic medication use or a cerebrovascular event that could affect the VF. Glaucomatous eyes were defined as having VF defects, as confirmed by at least two reliable VF examinations and the presence of a glaucomatous optic disc that showed diffuse or focal neural rim thinning, and/or RNFL defects. A glaucomatous VF defect was defined as a cluster of three or more points with a probability of <5% on the pattern deviation map in at least one hemifield, including at least one point with a probability of <1%; or a result “outside normal limits” in the glaucoma hemifield test; or a pattern standard deviation (PSD) with a probability of <5%. If both eyes were eligible for the study, one eye was selected randomly for inclusion. 

Optical Coherence Tomography

The SD-OCT imaging was performed using a Cirrus HD-OCT instrument software version 6.0 (Carl Zeiss Meditec, Inc.). For ganglion cell analysis, three-dimensional macular cube OCT data were obtained using the macular cube 512 × 128 scan mode. Images with signal strength of <6 or algorithm segmentation failure, and those with visible eye motion or blinking artifacts were considered poor quality and discarded. The algorithm identifies the RNFL (from the internal limiting membrane to the outer boundary of the RNFL) and the GCIPL layer (from the outer boundary of the RNFL to the outer boundary of the IPL; thus, a combination of the RGC layer and the IPL). The segmentation procedure operates in three dimensions and uses a graph-based algorithm to identify each layer. The average, minimum (lowest GCIPL thickness over a single meridian crossing the annulus), and sectoral (superotemporal, superior, superonasal, inferonasal, inferior, inferotemporal) thicknesses of the GCIPL were measured in an elliptical annulus (dimensions, vertical inner and outer radius of 0.5 and 2.0 mm, horizontal inner and outer radius of 0.6 and 2.4 mm, respectively). Details of the manner in which GCIPL thickness measurements are conducted have been reported previously. ^14,15 The superior hemifield GCIPL thickness was defined as the average of measurements in superonasal, superior, and superotemporal sectors, and the inferior hemifield GCIPL thickness was defined as the average of measurements in inferonasal, inferior, and inferotemporal sectors. 

Using SD-OCT, the pRNFL thickness was measured in the optic disc cube 200 × 200 scan mode, which scans a 6 × 6-mm square centered on the ONH to collect 200 × 200 axial scans containing 40,000 points. The Cirrus HD-OCT algorithms automatically identify the center of the optic disc and create an artificial B-scan in the shape of a circle with a 3.46 mm diameter around it, comprising 256 A-scans. The system calculates the pRNFL thickness at each point on the circle, and generates the overall average pRNFL thickness, pRNFL thickness of each quadrant (temporal, superior, nasal, and inferior quadrants) sector, and individual pRNFL thickness of all 12 clock-hour sectors. The macular pRNFL thickness was defined as the average of measurements in clock-hour segments 7 to 11. The temporal pRNFL thickness was defined as the average of the measurements in clock-hour segments 8 to 10. The superotemporal pRNFL thickness was defined as the average of measurements in clock-hour segments 10 and 11, and the inferotemporal pRNFL thickness was defined as the average of measurements in clock-hour segments 7 and 8. For a sectoral analysis, the pRNFL thickness of each clock-hour segments of 7, 8, 10 and 11 were included the calculation for correlation analysis between the GCIPL and pRNFL sectors, because the sectors correspond topographically to the regions in the GCIPL sectors (Fig. 1). The pRNFL thickness values in clock-hour segment 9 were excluded from the calculation for correlation analysis because this sector was not divided according to horizontal raphe. 

Figure 1

 

The VF of the Humphrey field analyzer Swedish interactive threshold algorithm 24-2 paradigm, GCIPL thickness measurements, and pRNFL thickness measurements by Cirrus HD-OCT of a right eye. (A) The central cluster MS was defined as the average of 12 central data points. The central cluster VF test points were grouped into four sectors, according to the structure–function correspondence map suggested by Garway-Heath et al. ¹⁸ Two test points located in the blind spot are marked with a cross. (B, C) Each of the regional VF test points and their corresponding GCIPL/pRNFL sectors are filled with the same color. (B) The number in each GCIPL sector represents the Pearson correlation coefficient between the corresponding VF sensitivity and the sectoral GCIPL. (C) The number in each pRNFL clock-hour sector represents the Pearson correlation coefficient between the corresponding VF sensitivity and the sectoral pRNFL. The temporal pRNFL thickness was defined as the average of the measurements in clock-hour segments 8 to 10, and is indicated by the solid yellow line. The macular pRNFL thickness was defined as the average of the measurements in clock-hour segments 7 to 11, and is indicated by the dashed blue line. The superotemporal (ST) pRNFL thickness was defined as the average of measurements in clock-hour segments 10 and 11, and is indicated by the solid red line. The inferotemporal (IT) pRNFL thickness was defined as the average of measurements in clock-hour segments 7 and 8 and is indicated by the solid green line. S, superior; SN, superonasal; I, inferior; IN, inferonasal.

VF Examination

Structure–function relationships were analyzed by comparing the mean deviation (MD), VF index (VFI), and corresponding mean sensitivity (MS) values measured by SAP, and the OCT parameter assessed by SD-OCT. The VF sensitivity was evaluated using a logarithmic decibel scale and a nonlogarithmic 1/L scale (where L is luminance measured in lamberts). The nonlogarithmic 1/L value at each tested location was calculated by dividing the value in decibels by 10, followed by conversion to the nonlogarithmic form: Retinal light sensitivity (decibels) = 10 × log₁₀ (1/L). Two test points located in the blind spot were excluded from the analysis. The superior hemiretinal MS was defined as the average of 26 test points in the superior hemifield, excluding the blind spot. Similarly, the inferior hemiretinal MS was defined as the average of 26 test points in the inferior hemifield, excluding the blind spot. ¹⁷ The central cluster MS, which was assumed to correspond topographically to the retina within 4.8 mm of the fovea, was defined as the average of 12 central data points. 

The superior center MS was defined as the average of the superior 6 test points of the 12 central cluster points, and the inferior center MS was defined as the average of the inferior 6 test points of the 12 central cluster points. Subsequently, the central cluster VF test points were grouped into four sectors topographically, according to the structure–function correspondence map suggested by Garway-Heath et al. ¹⁸ The superonasal center MS (SN center MS) was defined as the average of the MS in 4 superonasal points of the 12 central cluster points and the inferonasal center MS (IN center MS) was defined as the average MS in the 3 inferonasal points. The superotemporal center MS (ST center MS) was defined as the average of the MS in the 2 superotemporal points and the inferotemporal center MS (IT center MS) was defined as the average of the MS in 3 inferotemporal points (Fig. 1). 

Statistical Analyses

In all regression analyses, VF sensitivity was treated as the dependent variable, and the average/sectoral GCIPL thickness (GCIPLT) and pRNFL thickness (pRNFLT) parameters were independent variables. The correlations of the average/sectoral GCIPLT and pRNFLT values with VF sensitivity were evaluated using linear regression analyses. The relationship between VF sensitivity and the sectoral GCIPLT was analyzed for each of the following pairs: inferior center MS versus superior hemifield GCIPLT, superior center MS versus inferior hemifield GCIPLT, IN center MS versus ST GCIPLT, IT center MS versus S GCIPLT, IT center MS versus SN GCIPLT, ST center MS versus IN GCIPLT, ST center MS versus I GCIPLT, and SN center MS versus IT GCIPLT. 

To compare the associations between the corresponding VF MS and OCT measurements obtained using macular GCIPL and ONH scan modes, we assessed the significance of differences between any two correlation coefficients using the bootstrap method (1000 replicates). From the R means from 1000 samples reported for each relationship, a t-test was performed to test the null hypothesis that the R value between two models is equal. ¹⁹ All reported P values are 2-sided and differences at a level of P < 0.01 were considered to indicate statistical significance. The Statistical Package for the Social Sciences version 18.0 (SPSS, Inc., Chicago, IL) was used for all statistical analyses. 

Results

A total of 213 glaucomatous eyes was included in this cross-sectional study. Patient demographics are summarized in Table 1. 

Table 1

 

Demographic Data

	Glaucoma, N = 213
Age, y	53.7 ± 11.9
Sex, male:female	109:104
Visual field parameters
Mean deviation, dB	−6.3 ± 6.6
Pattern standard deviation, dB	7.1 ± 4.5
Visual field index, %	82.7 ± 19.5
Superior hemifield MS, dB	22.2 ± 8.4
Inferior hemifield MS, dB	26.2 ± 6.3
Central cluster MS, dB	25.9 ± 6.8
Superior cental MS, dB	23.1 ± 9.9
Inferior central MS, dB	28.6 ± 6.2
Peripapillary pRNFLT, μm
Average pRNFLT	75.8 ± 11.8
Superior pRNFLT	94.1 ± 20.5
Nasal pRNFLT	62.8 ± 8.5
Inferior pRNFLT	82.6 ± 22.2
Temporal pRNFLT	63.0 ± 11.6
Macular GCIPLT, μm
Average GCIPLT	71.7 ± 7.7
Minimum GCIPLT	60.6 ± 10.8
Superotemporal GCIPLT	71.8 ± 9.7
Superior GCIPLT	75.4 ± 9.3
Superonasal GCIPLT	79.2 ± 9.5
Inferonasal GCIPLT	74.0 ± 9.6
Inferior GCIPLT	65.9 ± 10.0
Inferotemporal GCIPLT	64.2 ± 10.8

Table 2 shows the correlation between the GCIPL and pRNFL thicknesses of the topographically corresponding sectors. Statistically significant correlations between the macular GCIPL thickness and corresponding pRNFL thickness were found in all GCIPL sectors (R = 0.534–0.807). 

Table 2

 

Pearson Correlation Coefficients Between Macular GCIPLT and pRNFLT

	R
Average GCIPLT vs. average pRNFLT	0.699 (<0.001)
Average GCIPLT vs. temporal pRNFLT	0.657 (<0.001)
Average GCIPLT vs. macular pRNFLT	0.700 (<0.001)
Superior GCIPLT vs. clock-hour 10 & 11 pRNFLT	0.630 (<0.001)
Inferior GCIPLT vs. clock-hour 7 & 8 pRNFLT	0.807 (<0.001)
ST GCIPLT vs. clock-hour 11 pRNFLT	0.579 (<0.001)
S GCIPLT vs. clock-hour 10 pRNFLT	0.681 (<0.001)
SN GCIPLT vs. clock-hour 10 pRNFLT	0.534 (<0.001)
IN GCIPLT vs. clock-hour 8 pRNFLT	0.631 (<0.001)
I GCIPLT vs. clock-hour 8 pRNFLT	0.759 (<0.001)
IT GCIPLT vs. clock-hour 7 pRNFLT	0.782 (<0.001)

 The numbers represent the Pearson correlation coefficients and P values (in parentheses).

Tables 3 and 4 show the relationships between VF sensitivity, expressed as decibel and 1/L scales, and GCIPL thickness/RNLF thickness. 

Table 3

 

Comparison of the Relationships Between Cirrus HD-OCT Thicknesses (GCIPLT, pRNFLT) and Global/Regional VF Sensitivities

	R		P Value
	GCIPLT	pRNFLT
Mean deviation, dB
Average GCIPLT vs. average pRNFLT	0.667 (<0.001)	0.689 (<0.001)	0.544
Visual field index, %
Average GCIPLT vs. average pRNFLT	0.676 (<0.001)	0.667 (<0.001)	0.806
Central cluster MS, dB
Average GCIPLT vs. temporal pRNFLT	0.710 (<0.001)	0.564 (<0.001)	<0.001*
Average GCIPLT vs. macular pRNFLT	0.710 (<0.001)	0.660 (<0.001)	0.158
Central cluster MS, 1/L
Average GCIPLT vs. temporal pRNFLT	0.612 (<0.001)	0.557 (<0.001)	0.206
Average GCIPLT vs. macular pRNFLT	0.612 (<0.001)	0.609 (<0.001)	0.940
Inferior center MS, dB
Superior hemifield GCIPLT vs. superotemporal pRNFLT	0.654 (<0.001)	0.584 (<0.001)	0.102
Inferior center MS, 1/L
Superior hemifield GCIPLT vs. superotemporal pRNFLT	0.542 (<0.001)	0.586 (<0.001)	0.169
Superior center MS, dB
Inferior hemifield GCIPLT vs. inferotemporal pRNFLT	0.747 (<0.001)	0.653 (<0.001)	0.001*
Superior center MS, 1/L
Inferior hemifield GCIPLT vs. inferotemporal pRNFLT	0.718 (<0.001)	0.654 (<0.001)	0.030

 The numbers represent the Pearson correlation coefficients (R) and P values (in parentheses). Comparison of the Pearson correlation coefficients (R) between the GCIPLT and pRNFLT was done using the Bootstrap method.

*   P < 0.01.

Table 4

 

Comparison of the Relationships Between Sectoral Cirrus HD-OCT Thicknesses (GCIPLT, pRNFLT) and Regional VF Sensitivities

	R		P Value
	GCIPLT	pRNFLT
IN center MS, dB
ST GCIPLT vs. clock-hour 11 pRNFLT	0.613 (<0.001)	0.553 (<0.001)	0.203
IN center MS, 1/L
ST GCIPLT vs. clock-hour 11 pRNFLT	0.575 (<0.001)	0.624 (<0.001)	0.286
IT center MS, dB
S GCIPLT vs. clock-hour 10 pRNFLT	0.570 (<0.001)	0.471 (<0.001)	0.026
IT center MS, 1/L
S GCIPLT vs. clock-hour 10 pRNFLT	0.452 (<0.001)	0.463 (<0.001)	0.814
IT center MS, dB
SN GCIPLT vs. clock-hour 10 pRNFLT	0.600 (<0.001)	0.471 (<0.001)	0.014
IT center MS, 1/L
SN GCIPLT vs. clock-hour 10 pRNFLT	0.479 (<0.001)	0.463 (<0.001)	0.776
ST center MS, dB
IN GCIPLT vs. clock-hour 8 pRNFLT	0.630 (<0.001)	0.578 (<0.001)	0.119
ST center MS, 1/L
IN GCIPLT vs. clock-hour 8 pRNFLT	0.542 (<0.001)	0.568 (<0.001)	0.289
ST center MS, dB
I GCIPLT vs. clock-hour 8 pRNFLT	0.666 (<0.001)	0.578 (<0.001)	0.007*
ST center MS, 1/L
I GCIPLT vs. clock-hour 8 pRNFLT	0.632 (<0.001)	0.568 (<0.001)	0.041
SN center MS, dB
IT GCIPLT vs. clock-hour 7 pRNFLT	0.669 (<0.001)	0.604 (<0.001)	0.026
SN center MS, 1/L
IT GCIPLT vs. clock-hour 7 pRNFLT	0.705 (<0.001)	0.664 (<0.001)	0.095

 Comparison of the Pearson correlation coefficients (R) between the GCIPLT and pRNFLT was done using the Bootstrap method. The numbers represent the Pearson correlation coefficients (R) and P values (in parentheses).

*   P < 0.01.

In Table 3, MD, VFI, and central cluster MS were correlated significantly with the average GCIPL thickness. In the comparative analysis, the association between central cluster MS and average GCIPL thickness was significantly stronger than that of the central cluster MS and temporal pRNFL thickness (in decibel scales, P < 0.001, Table 3). In the hemifield analysis, the strongest association was observed between the inferior hemifield GCIPL thickness and superior center MS pair (in decibel scales, R = 0.747, Table 3). In addition, the association between superior center MS and the inferior hemifield GCIPL thickness was significantly stronger than that of superior center MS and inferotemporal pRNFL thickness (in decibel scales, P = 0.001, Table 3). 

In Table 4, details of structure–function relationship, and comparison between each clock-hour segment of pRNFL thickness and corresponding sectoral GCIPL thickness is shown. Statistically significant correlations between regional VF MS (in the decibel and 1/L scales) and all sectoral GCIPL thicknesses were found (R = 0.452–0.705). In the sectoral comparative analysis, the association between corresponding VF sensitivities and the inferior GCIPL thickness was significantly stronger than that of corresponding VF sensitivities and pRNFL thickness using the decibel scale (P = 0.007, Table 4). In addition, the association between regional VF sensitivities and the superior, superonasal, and inferotemporal GCIPL thicknesses were more marked than those between corresponding VF sensitivities and pRNFL thicknesses, although the differences were not significant (P = 0.026, 0.014, and 0.026, Table 4). 

Discussion

We evaluated and compared the relationships between global and regional VF sensitivity, and the average/sectoral GCIPLT values and pRNFLT values in glaucomatous eyes. Mwanza et al. ¹⁶ evaluated the structure–function relationship between the VF mean deviation and GCIPL thickness. However, among all perimetry global indices, they used only the VF mean deviation. Although Raza et al. ²⁰ demonstrated that the association between local VF sensitivity and RGC+IPL thickness was strong, VF test points were grouped into 5 eccentricities, unlike our study. In our study, the global and regional VF sensitivities were correlated significantly with the average/sectoral GCIPLT for all macular GCIPLT values (Fig. 1). We also demonstrated that macular GCIPL analysis is more valuable than pRNFL analysis for understanding the structure–function relationship of the central retina in glaucoma patients with a central field defect. 

Previous studies have reported a good correlation between global and regional visual sensitivities, and structural measurements determined by various techniques. ^11,21–24 Structure–function relationships between pRNFL thickness and VF sensitivity in glaucomatous eyes generally tend to show more meaningful correlations in eyes with more advanced disease. ²² The correlation reported by Leung et al. ²² (R ² = 0.623, MD = −11.1 ± 7.74 dB) was much stronger than that reported by Bowd et al. ²³ (R ² = 0.38, MD = −3.0 ± 2.1 dB). The correlation between pRNFL thickness and VF sensitivity in a study by Nilforushan et al. ¹¹ (R ² = 0.24, MD = −1.3 ± 1.9 dB) was lower than both of these previously reported values. In our study, the MD was −6.30 ± 6.57 dB, and the R ² values of the highest correlation were consistent with the results of Bowd et al. ²³  

Axons of the RGCs located in the superotemporal and inferotemporal GCIPL sectors correspond to the superotemporal and inferotemporal regions, respectively, of the ONH. ²⁵ Therefore, in our study, we evaluated the relationship between SN center/IN center MS and the superotemporal/inferotemporal GCIPL thickness values. Additionally, the association between the ST/IT center MS and four other sectoral GCIPL thickness values was evaluated because these four GCIPL sectors were assumed to correspond to more central VF points, compared to the two temporal GCIPL sectors (Fig. 1). As shown in the Table 4, the association between SN center MS and the inferotemporal GCIPLT was strongest among the relationships between regional VF sensitivity and six sectoral GCIPLT. An excellent performance of inferotemporal GCIPL sectors was demonstrated in previous studies by Mwanza et al., ²⁶ who reported that the area under the curve (AUC) of the inferotemporal sector GCIPLT was significantly higher than that of the average GCIPLT (P = 0.049), and that among GCIPL parameters, the inferotemporal GCIPLT showed the best performance in the diagnosis of glaucoma. Previous studies also demonstrated a highly significant correlation between other inferotemporal sector OCT parameters, such as pRNFL thickness, and the corresponding VF MS. El Beltagi et al. ²⁷ reported linear relationships that were strongest between the inferotemporal pRNFL thickness and the superonasal VF (R ² = 0.57), and Bowd et al. ²¹ found that the structure–function relationship generally was strongest between the inferotemporal pRNFL thickness and the superonasal VF (R ² = 0.38). One reason for the strong correlation between inferotemporal sector structure and VF sensitivity may be related to the observation of RNFL defects in glaucoma most frequently at the inferotemporal meridian. ²⁸ On the other hand, in our study, the correlation between regional VF sensitivities and the inferior/inferonasal GCIPL sector was weaker than those with the inferotemporal GCIPL sector. This may be due to the fact that these regions contain more papillomacular nerve fiber bundles, which generally are the last nerve fibers to become severely involved in glaucoma, although parafoveal VF defects can occur during its early stages. 

The VFI was expressed as a percentage after age correction and a weighting procedure, and the relative importance of the central and paracentral test point locations are reflected more accurately by VFI. ²⁹ We investigated whether there was a structure–function relationship between the MD/VFI and GCIPL thickness. It might not be reasonable to compare the average GCIPL and pRNFL in terms of their relationship with the global VF index. However, the average GCIPL thickness was associated significantly with the VFI/MD; this association was comparable to that of the average pRNFL thickness, although approximately 50% of the RGCs are concentrated within 4.5 mm of the fovea. ³⁰ In addition, we analyzed data including only the 12 central data points. The central cluster MS in this study was defined as the average of 12 central data points. The associations of average GCIPL thickness with the central cluster VF MS (in the decibel scale) were significantly greater than that between corresponding VF sensitivity and temporal pRNFL thickness. As reported previously, ^14,22 the locally-weighted scatterplot smoothing (LOWESS) plot suggests a curvilinear relationship between VF sensitivity (using total deviation value [dB]) and GCIPL/pRNFL thickness. On the contrary, when VF sensitivity is expressed nonlogarithmically in the 1/L scale, the structure–function relationship is linear (Fig. 2). 

Figure 2

 

Scatter plots showing the correlations between the corresponding VF sensitivity (using total deviation) and Cirrus HD-OCT parameters (GCIPL thicknesses and pRNFL) in glaucomatous eyes. (A) Average GCIPL thickness versus central cluster sensitivity (dB). (B) Inferior hemifield GCIPL thickness versus superior center sensitivity (dB). (C) Temporal pRNFL thickness versus central cluster sensitivity (dB). (D) Inferotemporal pRNFL thickness versus superior center sensitivity (dB). (E) Average GCIPL thickness versus central cluster sensitivity (1/L). (F) Inferior hemifield GCIPL thickness versus superior center sensitivity (1/L). (G) Temporal pRNFL thickness versus central cluster sensitivity (1/L). (H) Inferotemporal pRNFL thickness versus superior center sensitivity (1/L).

In the hemifield analysis, the results of our study were in agreement with those of the earlier study in which the structure–function relationships with the GCC thickness was greater than that of macular pRNFL thickness in the superior macular hemifield (Table 3). ¹⁴ However, the analysis using sectoral GCIPL thicknesses provided us with more detailed informations about the structure–function relationship of the macular region (Table 4). In the sectoral comparative analysis, the associations between corresponding VF sensitivities and the superior, superonasal, inferior, and inferotemporal GCIPL thicknesses were stronger than those between regional VF sensitivities (in the decibel scale) and corresponding pRNFL thicknesses (Figs. 3, 4). There are a number of possible explanations for these results. First, the GCIPL is thicker in the macular region than in other peripheral retinal regions, whereas the temporal pRNFL is thinner than the superior and inferior pRNFL. The decrease in thinner temporal pRNFL thickness might not reflect a loss of RGC accurately. In addition, the anatomy of the macular area is less variable than that of other diagnostically important structures, such as the optic disc and pRNFL. Anatomic variations, such as extensive peripapillary atrophy (PPA) and optic disc tilt, can create measurement errors in ONH scans. ^31,32 The identification of the neuroretinal rim may be difficult in myopia patients with severe PPA, making the precise measurement of the pRNFL thickness challenging. Kim et al. ³² showed that the AUC for the pRNFL thickness determined by Cirrus OCT was decreased significantly in the glaucoma with PPA group compared to the glaucoma without PPA group. Furthermore, in addition to the RGC axons, the thickness of glial cells and blood vessels was measured with OCT as the pRNFL thickness. ³³ Hood et al. ³⁴ demonstrated that the pRNFL profile is correlated with the location of the peripapillary retinal vessels. Therefore, the distribution of the peripapillary retinal vessels may be the cause of the variability in the pRNFL thickness. These results suggest that GCIPL thickness measurements may be more valuable than temporal or macular pRNFL thickness for structural evaluation of the macular region in glaucomatous eyes. 

Figure 3

 

Representative case showing the relationship between the corresponding VF mean sensitivity and the sectoral GCIPL thickness in a glaucomatous eye with a superotemporal center VF defect. The red on the GCIPL macular sector map indicates regions with decreased GCIPL thicknesses outside normal limits. The slight decrease (red arrow) in the inferotemporal region of the TSNIT graph is suspicious. However, the pRNFL clock-hour sector map shows that all pRNFL thicknesses were within the normal range.

Figure 4

 

Two representative cases of glaucomatous eyes with a superonasal center VF defect. In case 1, the VFs demonstrate an early parafoveal scotoma, which is more severe in the VFs of case 2. Although the clock-hour 7 pRNFL thickness is similar between cases 1 and 2, the inferotemporal GCIPL is thinner in case 2 than in case 1. Functional loss in glaucomatous eyes is correlated strongly with the macular GCIPL thickness.

Theoretically, the nonlogarithmic 1/L scale might be more suitable than the logarithmic decibel scale for evaluation of the structure–function relationship. However, most of the correlations using decibel units were superior to those using 1/L units, especially in our GCIPL analysis, as noted in a previous study using GCC measurement. ¹⁹ This implies that the range of GCIPL thickness might be insufficient to reflect the functional change that is expressed nonlogarithmically in the 1/L scale. It is possible that the nonlogarithmic 1/L scale might be more suitable than the logarithmic decibel scale when the analysis is restricted to patients with mild or moderate VF defects, because the structure–function correlation was not evident in eyes with advanced glaucoma. ³⁵ Although there has been no clear explanation for the discrepancies between previous structure–function relationship studies, Leung et al. ²² showed that the relationship between the pRNFLT and VF sensitivity depends on the characteristics of the patient population, the type of imaging device, and the method of VF expression. 

Our study has limitations. First, VF examination using the central 10-2 program may be more appropriate for determination of the relationship between VF sensitivity and macular GCIPLT. It has been well known that 10-2 SAP tests provide more valuable information than 24-2 SAP test in the patient with parafoveal field defect. ^12,36–38 However, the 10-2 program is not a commonly-used VF algorithm and this study was a retrospective study. Therefore, we selected 12 central macular VF points using 24-2 SAP, which is the most widely used VF algorithm. In our study, using the 24-2 SAP test, we demonstrated that the association between the central cluster MS and average GCIPL thickness was significantly stronger than that of temporal pRNFL thickness. Further study using a 10-2 VF testing strategy will be needed to obtain more detailed information of the macular region. Second, for sectoral comparative analysis, we used the GCIPL and pRNFL sectors predefined in the commercial software supplied with the OCT machine. Therefore, the sectoral GCIPL area measured using Cirrus OCT GC/IPL scan mode does not perfectly match the sectoral pRNFL area topographically. Although we selected VF topographic sectors that are more appropriate for examining functional correlations with the pRNFL, rather than the macular region, the macular GCIPL thickness values provided more valuable information than pRNFL thickness values in terms of understanding the structure–function relationships. More extensive division of GCIPL and pRNFL measurements should provide a clearer picture of the structure–function relationship. Third, our study included only glaucoma patients. Further research, including suspected glaucoma patients, is needed to obtain more information about structure–function relationships. Fourth, this cross-sectional study did not demonstrate longitudinal structural and functional data. Therefore, a longitudinal study should be completed to correlate macular GCIPLT with functional declines over time in glaucomatous optic neuropathy. Finally, macular GCIPL analysis is not optimal for patients having any retinal disease, such as diabetic macula edema, epiretinal membrane, and age-related macular degeneration, that can affect macular thickness. 

In our current search for useful topographic relationships using Cirrus HD-OCT ganglion cell analysis algorithms, we found that global and sectoral macular GCIPL thickness measurements were associated significantly with the corresponding VF sensitivity in all GCIPL sectors, and that central functional loss in glaucomatous eyes was associated strongly with macular GCIPL thickness. The associations between superior center MS and the inferior hemifield GCIPL thickness were significantly greater than that of superior center MS and inferotemporal pRNFL thickness. Although GCIPL thickness is not a perfect parameter, the macular GCIPL thickness value provided by Cirrus HD-OCT may be more valuable than pRNFL thickness, particularly temporal pRNFL thickness, for the structural assessment of the macular region and for understanding the structure–function relationship. 

Acknowledgments

The authors alone are responsible for the content and writing of the paper. 

Disclosure: H.-Y. Shin, None; H.-Y.L. Park, None; K.I. Jung, None; C.K. Park, None