**Purpose**:
To determine if the relationship between retinal sensitivity and macular inner retinal layer thickness differs between primary open-angle glaucoma (POAG) with mild and advanced central visual field (VF) damage.

**Methods**:
One eye of 153 POAG patients was included. Using spectral-domain optical coherence tomography, we measured the average thickness of the macular ganglion cell–inner plexiform layers (GCIPLT) and the macular nerve fiber layer/GCIPL (ganglion cell complex [GCCT]) in a 0.9-mm-diameter ganglion cell displacement–adjusted circular area corresponding to the four central test points of the Humphrey Perimeter 24-2 program and correlated the results with the average retinal sensitivity (1/Lambert) at the corresponding test points, with adjustment for other confounding factors.

**Results**:
Ninety-three eyes had mild central and 60 eyes advanced central VF damage with an average total deviation (TD) of the four test points of greater than or equal to −4 decibels (dB) (mild group) and less than −4 dB (more severe group), respectively; the average mean deviations were −3.0 and −9.8 dB, respectively. In the mild group, the GCCT and GCIPLT were correlated significantly and positively with the average retinal sensitivity with partial regression coefficient of 0.007 and 0.005, respectively, and in the more severe group with partial regression coefficient of 0.019 = 0.007 + 0.012 (*P* = 0.007) and 0.010 = 0.005 + 0.005 (*P* = 0.078), respectively. The axial length and disc size were correlated with GCIPLT with marginal significance (*P* = 0.052 and *P* = 0.042).

**Conclusions**:
The relationship between the macular GCC and GCIPL thickness and retinal sensitivity at the corresponding retinal areas differed between POAG with mild and advanced central VF damage.

^{1}

^{2–14}Previous studies have agreed that the correlation between the OCT-measured thicknesses of the RGC-related retinal layers and corresponding SAP sensitivities is weak and insignificant when the SAP sensitivities are normal.

^{2–13}This finding has been explained by the fact that substantial RGC loss (i.e., at least 10% to approximately 20%) resulting from glaucomatous damage is needed to lower the measured SAP sensitivity to outside the normal range,

^{2,3,15}and it is reasonable to assume that the SAP sensitivities on the linear scale (1/Lambert = 10

^{dB value/10}) or the decibel (dB) scale decrease linearly with decreases in the thicknesses of the RGC-related retinal layers or the density of the RGCs on the dB scale, respectively, once the RGC loss exceeds the critical point. In studying the relationship between the SAP sensitivities and thicknesses of the RGC-related retinal layers in the corresponding retinal areas, the GCC or GCIPL thickness in the macular region where most RGCs are located is advantageous over adopting the cpRNFL thickness, as the axons comprising the cpRNFL originate from different retinal regions, and interindividual variation is believed to exist between the VF regions and the disc sectors.

^{16–19}

^{14}although the slope of the regression line was much flatter than those reported in glaucomatous eyes.

^{2,4–13}This finding suggested that the stage of glaucoma may also affect the relationship between the SAP sensitivities and thicknesses of the RGC-related layers in the corresponding retinal areas. In fact, Ajtony et al.

^{5}reported that correlation between cpRNFL thickness and mean deviation (MD) value given by HFA 30-2 program in primary open-angle glaucoma (POAG) eyes fit a curvilinear regression model rather than a linear regression model. To the best of the authors' knowledge, however, it has not been studied whether correlation between SAP sensitivities and GCC or GCIPL thickness in the corresponding retinal area depends on the disease stage or not.

^{20–25}

^{26}Eyes with POAG were grouped based on the mild (mild group) or advanced central VF damage (more severe group) according to the extent of damage at the four central test points of the HFA 24-2 program. If both eyes of a participant fulfilled the inclusion criteria, the eye with better spectral-domain (SD)-OCT image quality was enrolled.

^{2}area (512 × 128 pixels) centered on the fixation point over a period of approximately 2.5 seconds. To obtain accurately sized fundus images, the magnification was corrected according to the formula provided by the manufacturer based on the refractive error, corneal radius, and axial length. The correspondence of the fundus photographs and OCT images was confirmed automatically using an OCT projection image and localization of the major retinal vessels.

^{27}An experienced examiner (AT) confirmed the image segmentation. Data for the GCC were obtained with the following formula: (mRNFL + GCIPL).

^{20}

*y*= 1.29 × (

*x*+ 0.46)

^{0.67}, where

*y*indicates ganglion cell eccentricity and

*x*indicates cone eccentricity.

^{28}The dB values for the SAP sensitivity in the four central test points of the HFA 24-2 test program were anti-logged to obtain the sensitivity in the linear scale (1/Lambert = 10

^{0.1×dB value}),

^{2–4,6,8,11,29}of which the mean of the four central test points was calculated using the above four anti-logged dB values (LSAP

_{4 test points}). Age,

^{20–23,24,25}axial length or refraction,

^{21,23,25}and disc size

^{20}are external factors that reportedly can affect the OCT-measured thicknesses of the cpRNFL, GCC, or GCIPL. The laterality and gender were excluded from the external factors according to the result of our previous study.

^{14}The possible effect of these factors was considered by multiple linear regression analysis: where GCCT

_{4 test points}(GCIPLT

_{4 test points}) indicates the mean of the measured GCC (GCIPL) thickness in the four retinal areas in the mild and more severe groups, and GLA indicates a category variable that adopts 0 in the mild group and 1 in the more severe group. The effects of age, axial length, and disc size on the relationship between the GCCT

_{4 test points}(GCIPLT

_{4 test points}) and LSAP

_{4 test points}were assumed to be independent of disease stage. When the partial regression coefficients A

_{4}and A

_{5}both differed significantly from 0, it was thought that the GCCT

_{4 test points}(GCIPLT

_{4 test points}) changed along with the LSAP

_{4 test points}decrease, but the slope characterizing the relationship between them differed significantly between the mild and more severe groups after correction of other confounding factors. The data were analyzed using SPSS software (21.0J for Windows; SPSS Japan Inc., Tokyo, Japan).

_{4 test points}) of −4 dB or higher, and the more severe group had a TD

_{4 test points}less than −4 dB. The mild group included 93 eyes of 93 patients whose age, MD, and TD

_{4 test points}averaged 58.3 years, −3.0, and −0.8 dB, respectively (Table 1). The more severe group included 60 eyes of 60 patients whose age, MD, and TD

_{4 test points}averaged 59.4 years, −9.8, and −11.1 dB, respectively (Table 2). A simple regression analysis between GCCT

_{4 test points}(GCIPL

_{4 test points}) (μm) and LSAP

_{4 test points}(1/Lambert) without adjustment for other confounding factors yielded regression coefficient (slope) of 0.007 ± 0.002 (0.006 ± 0.001) (

*P*= 0.001[

*P*= 0.000]) in the mild group and the difference in the regression coefficient between the mild and more severe groups was 0.011 ± 0.004 (0.005 ± 0.003) (

*P*= 0.011[

*P*= 0.128]) with the

*R*

^{2}value of 0.399 (0.296) (Figs. 1 and 2).

**Table 1**

**Table 2**

**Figure 1**

**Figure 1**

**Figure 2**

**Figure 2**

_{4 test points}(μm), the LSAP

_{4 test points}(1/Lambert) were correlated significantly with partial regression coefficient of 0.007 ± 0.002 (

*P*= 0.002). The difference in the partial regression coefficient for the LSAP

_{4 test points}between the mild and more severe groups (A

_{5}) was 0.012 ± 0.004 (

*P*= 0.008). That is, when the GCCT

_{4 test points}(μm) was plotted against the LSAP

_{4 test points}(1/Lambert) after adjustment for other confounding factors, the slope (Δ GCCT

_{4 test points}/Δ LSAP

_{4 test points}) was approximately 2.7 times greater in the more severe than in the mild group (0.019 = 0.007 + 0.012 vs. 0.007). The

*R*

^{2}value for the regression was 0.433.

**Table 3**

**Table 4**

_{4 test points}(μm), the LSAP

_{4 test points}(1/Lambert) was correlated significantly with partial regression coefficients of 0.005 ± 0.002 (SE) (

*P*= 0.005) and the difference in the partial regression coefficient for the LSAP

_{4 test points}between the early and progressed groups (A

_{5}) was 0.005 ± 0.003 (

*P*= 0.078). That is, when GCIPLT

_{4 test points}(μm) was plotted against the LSAP

_{4 test points}(1/Lambert) after adjustment for other confounding factors, the slope (ΔGCIPLT

_{4 test points}/Δ LSAP

_{4 test points}) tended to be approximately two times greater in the more advanced than in the mild group (0.010 = 0.005 + 0.005 vs. 0.005). Further, the axial length and disc size was negatively (

*P*= 0.052 and

*P*= 0.043) correlated with the GCIPLT

_{4 test points}, respectively. The

*R*

^{2}value for regression was 0.347.

*P*< 0.008) and the difference in the partial regression coefficient between the both groups was 0.015 ± 0.006 (0.006 ± 0.0061), respectively (

*P*= 0.002 and 0.133). At the retinal areas corresponding to the two central test points in the lower hemifield, 136 eyes (136 cases) had mild central VF damage (mean TD ≥ −4 dB) and 17 eyes (17 cases) advanced central VF damage (mean TD <−4 dB). The partial regression coefficient of the LSAP for the GCCT (GCIPLT) was 0.009 ± 0.002 (0.006 ± 0.002) (

*P*≤ 0.001), whereas the differences in the partial regression coefficient between the both groups were 0.008 ± 0.011 (0.004 ± 0.009) (

*P*> 0.500).

*P*< 0.001) and the difference in the partial regression coefficient between the group with TD ≥ −4.0 dB (

*n*= 88) and that TD < −4.0 dB (

*n*= 65) group was 0.050 ± 0.016 (0.032 ± 0.011) (

*P*< 0.006). At the retinal areas corresponding to the inferior and superior nasal and inferior temporal test points, the numbers of eyes with TD ≥ −4.0 dB and those with TD < −4.0 dB were 121 to 143 and 10 to 32 of 153, respectively. The partial regression coefficient of the LSAP for the GCCT (GCIPLT) was 0.007 to approximately 0.009 (0.005 ∼ 0.007) (

*P*< 0.002), respectively, whereas the differences in the partial regression coefficient between the both groups were 0.009 to approximately 0.037 (0.001 ∼ 0.009) (

*P*= 0.147 ∼ 0.792), respectively.

^{30}first reported that the MD values were correlated significantly with the mean macular thickness measurements obtained using time-domain OCT in eyes with moderately advanced glaucoma. The GCIPL and GCC thicknesses also were reported to be correlated with the global or corresponding regional SAP sensitivity (in dB or the linear scale) in groups with only glaucomatous eyes or groups with glaucomatous and normal eyes.

^{6–9,11}Although correlation between cpRNFL, GCC, or GCIPL thickness and retinal sensitivity was the subject of many previous studies, to the best of our knowledge, the current study is the first to report a significant difference in the relationship between the GCC (GCIPL) thickness and SAP linear scale sensitivity in corresponding macular regions between eyes with mild and advanced central VF. That is, (ΔGCCT

_{4 test points}[ΔGCIPLT

_{4 test points}]/Δ LSAP

_{4 test points}) was approximately 2.7 (2.2) times greater, or given a same reduction in the retinal sensitivity, reduction in RGC-related inner retinal layer thickness will be approximately 2.5 times greater, in the macular area with advanced glaucomatous damage than in that with mild glaucomatous damage. Because many clinicians are focusing their attention to functional and structural test results obtained from the macular area of glaucoma eyes, we suppose that the current result would be clinically interesting. The higher significance (

*P*= 0.008 vs.

*P*= 0.078) for the intergroup difference in the GCCT/LSAP relationship than that in the GCIPLT/LSAP relationship is probably attributable to the fact that the GCCT included the macular RNFL thickness, in which a change parallels that in the GCIPLT. Using another SD-OCT instrument, Cho et al.

^{6}plotted the average GCC thickness against the mean sensitivity over the entire HFA 24-2 test field in the linear scale in glaucomatous eyes with a mean MD of −7.0 dB. If the Δaverage GCC thickness/Δmean sensitivity in the linear scale is approximated from the figure they presented, the value of approximately 0.025 is obtained, which is not far from that currently obtained for (ΔGCCT

_{4 test points}/ΔLSAP

_{4 test points}) after adjustment for other confounding factors, 0.019, in the current POAG eyes with advanced central VF damage with a mean MD of −9.8 dB.

^{5}reported that correlation between cpRNFL thickness and MD value in POAG eyes fit a curvilinear regression model rather than a linear regression model, which agrees with the current results obtained in the macular area. One possible explanation may be plasticity in the visual cortex or normal cerebral adaptation to chronic modifications in the visual input caused by slowly progressing glaucomatous damage.

^{31}

^{14}The corresponding GCCT

_{4 test points}(GCIPLT

_{4 test points}) in the mild group was 106.8 (79.0) μm, and that in the normal eyes

^{14}calculated under the same conditions was 121.5 (90.3) μm. The intergroup difference in GCCT

_{4 test points}(GCIPLT

_{4 test points}), 14.7 (11.3) μm, was highly significant (

*P*< 0.001) after adjustment for the intergroup differences in age, axial length, and SAP sensitivity, corresponding to approximately 35% of the dynamic range of the GCCT

_{4 test points}(GCIPLT

_{4 test points}) we approximated previously.

^{14}This finding agreed with the previous finding that a substantial number of RGCs was lost before manifest VF damage developed.

^{15}Further, it is obvious that the (ΔGCCT

_{4 test points}[GCIPLT

_{4 test points}]/Δ LSAP

_{4 test points}) value of the current mild group, 7 (5) μm/1000 (1/Lambert), cannot explain the above difference in the GCCT

_{4 test points}(GCIPLT

_{4 test points}), 14.7 (11.3) μm, and the difference in the LSAP

_{4 test points}, (2064 − 1625 ≈ 440 [1/Lambert]), which suggested that mild glaucomatous damage is associated not only with the reduced thickness of the RGC-related retinal layers (or number of RGCs), but also the change in the relationship between its thickness (or number of RGCs) and perceived visual sensitivity. Thus, it may not be adequate to analyze the difference between the normal eyes and those with mild glaucomatous damage using difference in the partial regression coefficient (slope) given by a same numerical formula. On the other hand, the difference in the GCCT

_{4 test points}(GCIPLT

_{4 test points}) between the mild and more severe groups, (16.1 = 106.8 − 90.7 [8.2 = 79.0 − 70.8] μm), and the LSAP

_{4 test points}, (809 = 1,625 − 816 [1/Lambert]) seems to be compatible with the (ΔGCCT

_{4 test points}[GCIPLT

_{4 test points}]/ΔLSAP

_{4 test points}) value of the more severe group, 19 (11) μm/1000 (1/Lambert).

_{4 test points}(GCIPLT

_{4 test points}) was correlated strongly with age,

^{14}whereas age was not correlated significantly in the current eyes. The effect of glaucoma on the GCCT

_{4 test points}(GCIPLT

_{4 test points}) may be much stronger than that of physiologic aging in eyes with POAG. In POAG eyes, the GCIPLT

_{4 test points}was negatively correlated with a partial regression coefficient of approximately 1.2 μm/mm to axial length, which agreed with the value reported by Mwanza et al.

^{25}in normal eyes, approximately 1.0 μm/mm. These findings suggested that the effect of refraction must be corrected when studying the relationship between the RGC-related retinal layer thickness and SAP sensitivity in both normal and POAG eyes. Disc size also was negatively correlated only with the GCIPLT

_{4 test points}in the current POAG eyes. Mwanza et al.

^{25}reported that the GCIPL was not correlated with disc size in normal eyes. Further, the relationship between the disc size and glaucomatous damage remains controversial.

^{32,33}Thus, it seems difficult to discuss the pathophysiologic relationship between the GCIPLT and disc size in POAG eyes based on the current results, and this issue deserves future study. The possibility of a cohort effect in this study may not be excluded.

_{4 test points}(GCIPLT

_{4 test points}) and SAP sensitivity in the linear scale. However, the current results suggested that the value of (ΔGCCT [ΔGCIPLT]/ΔLSAP) may be even greater in eyes with very advanced POAG, and fitting to an exponential curve may have to be considered. As discussed above, however, the value of (ΔGCCT

_{4 test points}[ΔGCIPLT

_{4 test points}]/Δ LSAP

_{4 test points}) derived from current multiple linear regression analysis could reasonably explain the difference in GCCT

_{4 test points}(GCIPLT

_{4 test points}) and LSAP

_{4 test points}between the mild and more severe POAG group. This result may indicate that the linear regression model was an acceptable approximation as far as the current subjects were concerned. Forth, structure-function relationship in the macular area should be better studied adopting HFA 10-2 test results. What was reported in the current communication is the relationship at the most central four test points of HFA 24-2 test program. Future studies using HAF 10-2 test results should yield more information on the disease stage–dependent difference in the structure-function relationship in the macular area.

**M. Araie**, Alcon Japan (C, R), Allergan (C), Bosch-Lomb (C), Carl Zeiss Meditec (R), Kowa (C, R), Otsuka (R), Pfizer Japan (C, R), Santen (C, R), Senju (C, R), Topcon (C, R) P;

**H. Murata**, None;

**A. Iwase**, Alcon Japan (R), Carl Zeiss Meditec (R), Kowa (R), Otsuka (R), Pfizer Japan (R), Santen (R), Senju (R), Topcon (R), P;

**M. Hangai**, Nidek (C, F, R);

**K. Sugiyama**, None;

**N. Yoshimura**, Nidek (C, F, R)

*. 2000; 15: 221–233.*

*Sem Ophthalmol**. 2007; 26: 688–710.*

*Prog Retin Eye Res**. 2010; 29: 249–271.*

*Prog Retin Eye Res**. 2009; 50: 4254–4266.*

*Invest Ophthalmol Vis Sci**. 2007; 48: 258–263.*

*Invest Ophthalmol Vis Sci**. 2010; 51: 6401–6407.*

*Invest Ophthalmol Vis Sci**. 2010; 51: 4646–4651.*

*Invest Ophthalmol Vis Sci**. 2011; 129: 864–871.*

*Arch Ophthalmol**. 2011; 129: 1529–1536.*

*Arch Ophthalmol**. 2012; 21: 95–101.*

*J Glaucoma**. 2013; 54: 7344–7353.*

*Invest Ophthalmol Vis Sci**. 2012; 96: 47–52.*

*Br J Ophthalmol**. 2014; 157: 953–959.*

*Am J Ophthalmol**. 2014; 55: 7199–7205.*

*Invest Ophthalmol Vis Sci**. 1989; 107: 453–464.*

*Am J Ophthalmol**. 1982; 89: 255–267.*

*Ophthalmology**. 1990; 68: 568–574.*

*Acta Ophthalmol**. 2000; 107: 1809–1815.*

*Ophthalmology**. 2008; 49: 3018–3025.*

*Invest Ophthalmol Vis Sci**. 2010; 128: 1420–1426.*

*Arch Ophthalmol**. 2011; 52: 8769–8779.*

*Invest Ophthalmol Vis Sci**. 2011; 118: 2403–2408.*

*Ophthalmology**. 2012; 130: 312–318.*

*Arch Ophthalmol**. 2010; 51: 465–467.*

*Invest Ophthalmol Vis Sci**. 2011; 52: 7872–7879.*

*Invest Ophthalmol Vis Sci**. St. Louis: Mosby Inc.; 1999: 152–153.*

*Automated Static Perimetry**. 2010; 18: 21293–21307.*

*Opt Express**. 1999; 237: 1014–1023.*

*Graefes Arch Clin Exp Ophthalmol**. 2004; 45: 466–472.*

*Invest Ophthalmol Vis Sci**. 2003; 121: 41–46.*

*Arch Ophthalmol**. 1999; 19: 84–88.*

*J Neuroophthalmol**. 1999; 8: 347–352.*

*J Glaucoma**. 2003; 135: 226–228.*

*Am J Ophthalmol*