Structure–Function Relationships in Normal and Glaucomatous Eyes Determined by Time- and Spectral-Domain Optical Coherence Tomography

Abstract

Purpose.: To compare the relationships between retinal mean sensitivity (MS) and retinal nerve fiber layer (RNFL) thickness, as measured by time-domain (TD) and spectral-domain (SD) optical coherence tomography (OCT).

Methods.: Recruited subjects were divided into normal, glaucoma suspect, and glaucoma groups. RNFL thickness was measured with TD- and SD-OCT, and MS was assessed with visual field perimetry and expressed in decibels and 1/L, where L is luminance in lamberts. The relationship between SUPERIOR MS and INFERIOR RNFL thickness (clock-hour segments 5, 6, 7, and 8) and that between INFERIOR MS and SUPERIOR RNFL thickness (clock-hour segments 10, 11, 12, 1, 2, and 3) were correlated by linear and logarithmic regression analyses. Pearson's correlation coefficients (R), for both OCTs were compared by using Hotelling's t-test.

Results.: Ninety-five eyes of 76 subjects were prospectively included. Twenty-five eyes were classified as normal, 25 with glaucoma suspect, and 45 with glaucoma. In normal and glaucoma suspect eyes, there were no significant relationships between MS and RNFL thickness. In glaucomatous eyes, the associations between MS and RNFL thickness were R = 0.31 to 0.57 with TD-OCT and R = 0.47 to 0.66 with SD-OCT, and the correlation of SUPERIOR RNFL thickness with INFERIOR MS was significantly better with SD-OCT than with TD-OCT in both linear and logarithmic regression models.

Conclusions.: The results showed that, in mild-to-moderate glaucoma, SD-OCT offers an improved structure–function correlation compared with TD-OCT, when applied to the detection of INFERIOR MS and SUPERIOR RNFL defects.

In investigating structure–function relationships in glaucomatous eyes, several attempts have been made to correlate peripapillary retinal nerve fiber layer (RNFL) thickness distribution with particular visual field (VF) regions by using standard automated perimetry. ^{1 –4} Garway-Heath et al. ³ divided the optic disc into six sectors and reported the most detailed associations available to date between regions of the optic nerve head relevant to RNFL defects and VF test points, in patients with normal-tension glaucoma. ³ Although the approach is valuable, investigators in the cited study used only monochromatic photography as a qualitative method. Thus, Ferreras et al. ⁴ used time-domain (TD) optical coherence tomography (OCT; Stratus OCT, Carl Zeiss Meditec, Inc., Dublin, CA) as a quantitative tool to seek relationships between structure and function in glaucomatous eyes. 

OCT, first introduced in 1991 by Huang et al., ⁵ is a high-resolution, cross-sectional imaging instrument that assesses retinal architecture by using low-coherence, near-infrared light. The current third-generation instrument, Stratus OCT, provides quantitative and reproducible measurements of RNFL thickness ^6,7 and can detect RNFL thickness changes in glaucoma associated with VF loss. ^8,9 The most recent device, Cirrus HD-OCT (Carl Zeiss Meditec), has a spectral-domain (SD) feature with an axial image resolution of 5 μm and an image speed of 27,000 A-scans per second, thus providing better RNFL measurement reproducibility, ^{10 –12} with higher sensitivity ^13,14 and specificity, than the TD-OCT. ¹⁴ Because the two OCTs measure RNFL thickness according to different principles (SD versus TD), it is possible that the anatomic RNFL measurements of the two instruments bear different relationships to functional visual sensitivity. 

The purpose of this study was to evaluate and compare the structure–function relationships between retinal mean sensitivity (MS), as measured by standard automated perimetry, and RNFL thickness, as assessed by TD- and SD-OCT, in eyes that had no glaucoma, glaucoma suspect, or glaucoma, in a single ethnic population. In addition, we investigated whether linear or logarithmic models better fitted the observed structure–function relationships when RNFL thickness was expressed in linear form and MS was represented in either a linear or logarithmic model. 

Materials and Methods

Subjects and Measurement Protocol

The design of this study adhered to the tenets of the Declaration of Helsinki and was approved by the Institutional Review Board of HanGil Eye Hospital. Written, informed consent was obtained from each subject. From April 2008 to October 2008, subjects were prospectively enrolled from HanGil Eye Hospital, Incheon, Republic of Korea, and Seoul National University Hospital, Seoul, Republic of Korea. Each subject underwent a slit lamp examination; intraocular pressure (IOP) measurement with a Goldmann applanation tonometer (GAT), stereoscopic optic nerve photography and red-free RNFL photography, VF examination (Humphrey Field Analyzer/Carl Zeiss Meditec), TD-OCT (Stratus OCT), and SD-OCT (Cirrus HD-OCT) measurements. 

For inclusion in the study, subjects had to meet the following criteria: best corrected visual acuity of 20/40 or better and refractive spherical error within the range of −6.00 to +3.00 D or a cylinder within the range of ± 2.50 D. Subjects were excluded on the basis of any of the following criteria: (1) a history of any retinal disease, including diabetic or hypertensive retinopathy; (2) a history of eye trauma or surgery including trabeculectomy or Ahmed glaucoma valve implantation, with the exception of uncomplicated cataract surgery; (3) other optic nerve disease except for glaucoma; (4) a history of amblyopia; (5) an abnormal pupil on examination; (6) a history of use of miotics or other medications that can affect pupil size; (7) a history of systemic medication use that can affect the VF; or, (8) a history of a cerebrovascular event. 

Eyes of subjects were classified into three groups: normal, glaucoma suspect, and glaucoma. Normal eyes had an IOP of ≤21 mm Hg according to GAT; no glaucomatous optic neuropathy (GON), which was defined as diffuse or focal thinning (such as notching, or acquired pitting of the optic nerve) of the neuroretinal rim on stereoscopic optic nerve photography; no RNFL defect on red-free RNFL photography; and no glaucomatous VF defect (GVFD) on Humphrey VF examination. Eyes with glaucoma suspect had an IOP of >21 mm Hg and/or GON and/or RNFL defect, but no GVFD. Eyes with glaucoma had GON and RNFL defect with corresponding GVFD regardless of IOP. The criteria for GVFD were as follows: (1) a cluster of three or more points with a probability of <5%, with at least one point with a probability of <1%, on the same side of the horizontal meridian in the pattern deviation probability map; (2) a result “outside normal limits” in the glaucoma hemifield test; and (3) a pattern standard deviation with a probability of <5%. A reliable VF was defined as having a fixation loss of <20% and false-positive and -negative errors of <25%. 

Structure–function relationships were analyzed by comparing MS data (raw thresholds) measured by the Humphrey VF analyzer (HVF) with the Swedish interactive threshold algorithm (SITA) standard 24-2 program and RNFL thickness assessed by TD- and SD-OCT. In all glaucomatous eyes, at least two reliable HVF examinations were performed to minimize false positives. Retinal sensitivity was expressed in two forms: a logarithmic decibel scale and a nonlogarithmic 1/L scale (where L is luminance measured in lamberts). The differential light sensitivity (DLS) at each tested location can be written as: DLS (decibels) = 10 × log₁₀(1/L). 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. Two test points adjacent to the blind spot on the total sensitivity map were excluded from the analysis. The SUPERIOR MS was defined as the average of 26 data points from the superior hemifield, excluding the blind spot, and the INFERIOR MS was the average of 26 data points from the INFERIOR hemifield, again excluding the blind spot. GLOBAL MS was defined as the average of 52 data points from both hemifields, excluding the two blind spots. RNFL thickness was measured by both TD- and SD-OCT. With the TD-OCT, RNFL thickness was assessed by using the fast RNFL scan mode, which acquires 256 data points by averaging the results of three sequential circular scans, each of 3.4-mm diameter, centered on the optic nerve head. With the SD-OCT, RNFL thickness was measured in the optic disc cube 200 × 200 scan mode, which performs raster scanning in a 6 × 6-mm square centered on the optic nerve head. A scan consists of 200 frames of horizontal linear B-scans with 200 A-scans in each B-scan. RNFL thicknesses were measured overall, in four quadrants (temporal, superior, nasal, and inferior), and in 12 clock-hour sectors. Good-quality scan data, obtained with either type of OCT, were defined as those in scans well-centered on the optic nerve head with signal strengths over 5. Also images with algorithm failure were discarded. All data were aligned with respect to the orientation of the right eye, and so clock-hours 3 and 9 of a peripapillary scan represented the nasal and temporal sides, respectively, of the optic disc in both eyes. Specifically, we constructed VF sensitivity maps reflecting the SUPERIOR or INFERIOR RNFL thickness data acquired by TD- or SD-OCT as suggested by Ferreras et al., ⁴ excluding RNFL thickness measurements from clock-hour segments 4 and 9, because the cited studies showed that neither of these segment thicknesses correlated with VF sensitivity. Thus, INFERIOR RNFL thickness was defined as the average of data from clock-hour segments 5, 6, 7, and 8, and SUPERIOR RNFL thickness as the average of measurements in clock-hours 10, 11, 12, 1, 2, and 3. GLOBAL RNFL thickness was the average of data from all clock-hour segments except for segments 4 and 9 (Fig. 1). 

Figure 1.

 

Structure–function map between SUPERIOR RNFL and INFERIOR RNFL from TD- and SD-OCT and corresponding MS from visual field perimetry with 24-2 SITA program for a right eye, based on the method of Fererras et al. ⁴ (A) Sectors of the peripapillary RNFL except clock-hour segments 4 and 9 are divided into SUPERIOR (clock-hour segments 10, 11, 12, 1, 2, and 3) and INFERIOR RNFL (clock-hour segments 5, 6, 7, and 8). (B) Corresponding MS was the average of 26 points from each hemifield in a total sensitivity map in which each of the test points were numbered. Corresponding sectors were gray-scaled.

Statistical Analysis

Baseline demographics and clinical characteristics of normal, suspect, and glaucomatous eyes were compared by analysis of variance (ANOVA). The correlation of GLOBAL RNFL thickness with GLOBAL MS was performed with both linear (y = ax + b) and logarithmic [y = a ln (x) + b] regression analyses. Also, the relationship between SUPERIOR MS and INFERIOR RNFL thickness (clock-hour segments 5, 6, 7, and 8) and that between INFERIOR MS and SUPERIOR RNFL thickness (clock-hour segments 10, 11, 12, 1, 2, and 3) were evaluated in the same fashion. In any particular regression model, the degree of correlation between two variables was expressed as Pearson's correlation coefficient, R value. To assess the strength of association between TD- and SD-OCT data, we compared R values for both OCTs by using Hotelling's t-test, the statistical method that compares the relationship between variable z and variable x with the association between variable z and variable y in variables x, y, z. In all analyses, differences at a level of P < 0.05 were considered to be statistically significant. 

Results

A total of 130 eyes of 90 subjects were prospectively assessed. Ninety-five eyes of 76 subjects met the inclusion criteria and had a transparent crystalline lens or mild lens opacity according to the Lens Opacities Classification System III or pseudophakia. Among these 95 eyes, 25 were classified as normal, 25 as glaucoma suspect, and 45 as glaucomatous. Men and women were equal in number, with 38 subjects of either sex, and all were Korean. In the three groups, the mean ± standard deviation of age was 54.0 ± 10.5, 49.2 ± 14.8, and 56.2 ± 13.6 years, respectively; there were no significant differences in age or male-to-female ratio among the three groups. There were statistically significant differences among the three groups in IOP and all VF parameters, including mean deviation, pattern standard deviation, GLOBAL MS, SUPERIOR MS, and INFERIOR MS (all in decibels; Table 1). 

Table 1.

 

Descriptive Statistics for Demographic and Visual Field Parameters

	Total (n = 95)	Normal (n = 25)	Glaucoma Suspect (n = 25)	Glaucoma (n = 45)	P
Age, y	53.7 ± 13.3	54.0 ± 10.5	49.2 ± 14.8	56.2 ± 13.6	0.172
Sex, M:F	38:38	9:13	8:12	21:13	0.182
IOP, mm Hg	14.8 ± 3.6	13.6 ± 2.7	17 ± 4.4	14.2 ± 3.0	0.001*
VF parameters and retinal sensitivity, dB
MD	−3.5 ± 4.3	−0.9 ± 2.3	−1.7 ± 1.7	−6.0 ± 4.9	<0.001*
PSD	3.7 ± 3.4	1.8 ± 0.6	2.0 ± 0.8	5.8 ± 4.0	<0.001*
GLOBAL MS	26.8 ± 4.6	29.4 ± 2.7	28.8 ± 2.3	24.2 ± 5.0	<0.001*
SUPERIOR MS	25.7 ± 5.8	29.0 ± 2.9	28.2 ± 2.5	22.5 ± 6.6	<0.001*
INFERIOR MS	27.8 ± 4.3	29.9 ± 2.6	29.5 ± 2.1	25.8 ± 5.0	<0.001*

 Data are expressed as the mean ± SD.

*   P < 0.05, ANOVA test.

The mean overall RNFL thicknesses of normal, glaucoma suspect, and glaucomatous eyes were 109.8 ± 10.0, 99.4 ± 11.7, and 82.1 ± 14.0 μm, as measured by TD-OCT, and 95.0 ± 7.3, 89.7 ± 10.0, and 73.2 ± 9.9 μm, as measured by SD-OCT, respectively. RNFL thickness as measured by TD-OCT was significantly greater than that assessed by SD-OCT in all sectors of normal eyes, in all sectors except for clock-hour segments 3, 11, and 12 in the glaucoma suspect group, and in all sectors except for clock-hour segment 3 in the glaucoma group. RNFL thickness measured by TD-OCT correlated well with that measured by SD-OCT in all sectors in both normal and glaucoma suspect eyes, and in all sectors except for clock-hour segment 3 in glaucomatous eyes (Table 2). 

Table 2.

 

Comparison and Correlation of RNFL Thickness Measurements

OCT Parameters	Normal (n = 25)				Glaucoma Suspect (n = 25)				Glaucoma (n = 45)
	TD	SD	t-test (P)	R (P)	TD	SD	t-test (P)	R (P)	TD	SD	t-test (P)	R (P)
Overall	109.8 ± 10.0	95.0 ± 7.3	<0.001*	0.879 (<0.001)†	99.4 ± 11.7	89.7 ± 10.0	<0.001*	0.850 (<0.001)†	82.1 ± 14.0	73.2 ± 9.9	<0.001*	0.842 (<0.001)†
Temporal	82.1 ± 15.0	71.3 ± 13.2	<0.001*	0.909 (<0.001)†	77.9 ± 15.7	68.3 ± 14.9	<0.001*	0.832 (<0.001)†	69.0 ± 17.9	60.8 ± 15.3	<0.001*	0.952 (<0.001)†
superior	132.6 ± 12.2	119.6 ± 9.3	<0.001*	0.670 (<0.001)†	119.6 ± 25.5	111.6 ± 17.5	0.021*	0.783 (<0.001)†	100.8 ± 22.3	89.1 ± 20.6	<0.001*	0.766 (<0.001)†
Nasal	85.5 ± 15.8	69.4 ± 9.5	<0.001*	0.822 (<0.001)†	72.6 ± 15.8	66.3 ± 9.3	0.015*	0.658 (<0.001)†	66.5 ± 15.0	62.3 ± 7.8	0.037*	0.502 (<0.001)†
inferior	138.7 ± 19.5	119.8 ± 15.6	<0.001*	0.907 (<0.001)†	127.5 ± 16.0	112.3 ± 16.4	<0.001*	0.781 (<0.001)†	92.1 ± 23.9	80.8 ± 20.7	<0.001*	0.938 (<0.001)†
9 h	64.2 ± 12.5	54.7 ± 9.9	<0.001*	0.815 (<0.001)†	65.3 ± 13.4	55.6 ± 11.7	<0.001*	0.616 (0.001)†	58.2 ± 13.0	51.5 ± 12.9	<0.001*	0.720 (<0.001)†
10 h	98.2 ± 21.7	85.4 ± 20.6	<0.001*	0.894 (<0.001)†	93.1 ± 20.9	80.6 ± 16.6	<0.001*	0.750 (<0.001)†	77.9 ± 28.4	67.0 ± 21.9	<0.001*	0.926 (<0.001)†
11 h	142.6 ± 22.4	131.5 ± 23.6	<0.001*	0.822 (<0.001)†	128.7 ± 21.5	126.4 ± 20.7	0.599	0.469 (0.018)†	104.5 ± 31.5	95.2 ± 33.0	<0.001*	0.884 (<0.001)†
12 h	128.9 ± 22.1	117.1 ± 21.3	<0.001*	0.862 (<0.001)†	120.7 ± 37.9	114.2 ± 32.5	0.084	0.883 (<0.001)†	99.5 ± 28.0	86.5 ± 27.8	<0.001*	0.799 (<0.001)†
1 h	126.7 ± 16.4	110.3 ± 22.5	<0.001*	0.820 (<0.001)†	109.6 ± 31.1	94.3 ± 20.9	0.001*	0.800 (<0.001)†	98.3 ± 23.1	85.5 ± 19.3	<0.001*	0.729 (<0.001)†
2 h	101.6 ± 18.3	85.0 ± 16.5	<0.001*	0.714 (<0.001)†	84.6 ± 24.2	77.1 ± 14.0	0.030*	0.761 (<0.001)†	76.0 ± 21.1	68.8 ± 10.8	0.003*	0.736 (<0.001)†
3 h	72.3 ± 17.2	58.8 ± 8.2	<0.001*	0.850 (<0.001)†	62.0 ± 15.7	57.8 ± 11.3	0.132	0.537 (0.006)†	57.0 ± 13.7	57.8 ± 9.1	0.720	0.244 (0.106)
4 h	82.7 ± 16.4	64.7 ± 9.7	<0.001*	0.742 (<0.001)†	71.1 ± 15.2	63.8 ± 11.9	0.021*	0.419 (0.037)†	66.5 ± 16.9	60.2 ± 9.9	0.005*	0.534 (<0.001)†
5 h	114.6 ± 22.5	92.9 ± 18.8	<0.001*	0.939 (<0.001)†	102.7 ± 19.1	87.0 ± 17.4	<0.001*	0.777 (<0.001)†	86.0 ± 19.6	72.7 ± 14.2	<0.001*	0.811 (<0.001)†
6 h	148.3 ± 28.6	127.7 ± 26.0	<0.001*	0.840 (<0.001)†	138.6 ± 23.3	119.3 ± 23.3	<0.001*	0.621 (0.001)†	92.1 ± 27.0	80.7 ± 25.9	<0.001*	0.922 (<0.001)†
7 h	153.2 ± 20.1	138.8 ± 20.5	<0.001*	0.914 (<0.001)†	141.2 ± 22.3	130.6 ± 24.6	0.001*	0.820 (<0.001)†	98.2 ± 37.2	89.1 ± 31.6	<0.001*	0.948 (<0.001)†
8 h	84.1 ± 15.9	73.3 ± 17.3	<0.001*	0.938 (<0.001)†	75.4 ± 17.4	69.0 ± 20.0	0.005*	0.858 (<0.001)†	71.1 ± 20.3	64.1 ± 18.6	<0.001*	0.810 (<0.001)†

 Measurements were taken overall and in each of four quadrants and in each of 12 clock-hour segments determined by TD- and SD-OCT in the normal, glaucoma suspect, and glaucoma groups by paired t-test and Pearson correlation analysis.

*   P < 0.05, paired t-test.

†   P < 0.05, Pearson correlation analysis.

The mean GLOBAL RNFL thicknesses of normal, glaucoma suspect, and glaucomatous eyes were 117.1 ± 10.4, 105.7 ± 13.1, and 86.0 ± 15.0 μm as measured by TD-OCT, and 102.1 ± 8.0, 95.6 ± 10.9, and 76.7 ± 11.0 μm as assessed by SD-OCT, respectively. RNFL thickness measured by TD-OCT was significantly greater than that measured by SD-OCT in the GLOBAL, SUPERIOR, and INFERIOR sectors in all groups. RNFL thickness measured by TD-OCT was well correlated with that assessed by SD-OCT in the GLOBAL, SUPERIOR, and INFERIOR sectors of all groups (Table 3). 

Table 3.

 

Comparison and Correlation of GLOBAL, SUPERIOR, and INFERIOR RNFL Thickness Measurements

OCT Parameters	Normal (n = 25)				Glaucoma Suspect (n = 25)				Glaucoma (n = 45)
	TD	SD	t-test (P)	R (P)	TD	SD	t-test (P)	R (P)	TD	SD	t-test (P)	R (P)
GLOBAL	117.1 ± 10.4	102.1 ± 8.0	<0.001*	0.883 (<0.001)†	105.7 ± 13.1	95.6 ± 10.9	<0.001*	0.873 (<0.001)†	86.0 ± 15.0	76.7 ± 11.0	<0.001*	0.851 (<0.001)†
SUPERIOR	111.7 ± 9.8	98.0 ± 6.9	<0.001*	0.702 (<0.001)†	99.8 ± 16.0	91.7 ± 10.4	<0.001*	0.811 (<0.001)†	85.5 ± 16.7	76.8 ± 13.0	<0.001*	0.782 (<0.001)†
INFERIOR	125.0 ± 15.0	108.2 ± 11.6	<0.001*	0.910 (<0.001)†	114.5 ± 12.9	101.5 ± 13.7	<0.001*	0.811 (<0.001)†	86.8 ± 21.2	76.6 ± 17.4	<0.001*	0.941 (<0.001)†

 Measurements were determined by TD OCT and SD OCT in normal, glaucoma suspect, and glaucoma groups by paired t-test and Pearson correlation analysis.

*   P < 0.05, paired t-test.

†   P < 0.05, Pearson correlation analysis.

In glaucomatous eyes, relationships between MS expressed in decibels and 1/L, including GLOBAL, SUPERIOR, and INFERIOR data, and the corresponding RNFL thickness measurements, were statistically significant when analyzed by TD-OCT (R = 0.31–0.57, P ≤ 0.001–0.038) and SD-OCT (R = 0.47–0.66, P ≤ 0.001). However, in normal or glaucoma suspect eyes, there were no correlations between MS measurements, expressed in either decibels or 1/L, and RNFL thickness values in any of GLOBAL, SUPERIOR, or INFERIOR sectors obtained with either TD-OCT (P = 0.137–0.996) or SD-OCT (P = 0.446–0.924). In glaucomatous eyes, the correlation of SUPERIOR RNFL thickness with INFERIOR MS, expressed in decibels and 1/L, was significantly better when SD-OCT data were used, compared with TD-OCT values, in both linear and logarithmic regression models (P = 0.000–0.039). The correlation of GLOBAL RNFL thickness with GLOBAL MS expressed in either decibels or 1/L was significantly better with SD-OCT data than with TD-OCT data, in both the linear and the logarithmic regression models (P = 0.021–0.044), with the exception of GLOBAL MS expressed as 1/L in the logarithmic regression model (P = 0.060). However, the association of INFERIOR RNFL thickness with SUPERIOR MS expressed in either decibels or 1/L was not significantly better with SD-OCT versus TD-OCT data in either the linear or the logarithmic regression models (P = 0.247–0.906; Table 4, Fig. 2). 

Table 4.

 

Relationship between MS and RNFL thickness measured by TD-OCT and SD-OCT in Normal, Glaucoma Suspect, Glaucomatous Groups

OCT Parameters	HVF Parameters	Normal (n = 25)			Glaucoma Suspect (n = 25)			Glaucoma (n = 45)
		TD R (P)	SD R (P)	P	TD R (P)	SD R (P)	P	TD R (P)	SD R (P)	P
GLOBAL	GLOBAL MS (dB), Linear	0.25 (0.234)	0.07 (0.738)	—	0.04 (0.837)	0.03 (0.875)	—	0.37 (0.013)*	0.52 (<0.001)*	0.039†
	GLOBAL MS (dB), Log	0.24 (0.242)	0.06 (0.770)	—	0.05 (0.822)	0.04 (0.866)	—	0.36 (0.014)*	0.52 (<0.001)*	0.044†
	GLOBAL MS (I/L), Linear	0.31 (0.137)	0.15 (0.471)	—	−0.01 (0.952)	−0.04 (0.858)	—	0.31 (0.038)*	0.48 (<0.001)*	0.021†
	GLOBAL MS (I/L), Log	0.31 (0.137)	0.14 (0.491)	—	−0.001 (0.996)	−0.03 (0.885)	—	0.32 (0.031)*	0.47 (0.001)*	0.060
SUPERIOR	INFERIOR MS (dB), Linear	0.23 (0.263)	0.10 (0.645)	—	0.15 (0.461)	0.07 (0.747)	—	0.39 (0.008)*	0.64 (<0.001)*	0.003†
	INFERIOR MS (dB), Log	0.23 (0.268)	0.09 (0.684)	—	0.19 (0.368)	0.08 (0.694)	—	0.40 (0.007)*	0.66 (<0.001)*	0.005†
	INFERIOR MS (I/L), Linear	0.28 (0.175)	0.16 (0.446)	—	0.05 (0.809)	−0.05 (0.821)	—	0.37 (0.012)*	0.59 (<0.001)*	0.011†
	INFERIOR MS (I/L), Log	0.28 (0.173)	0.15 (0.471)	—	0.09 (0.685)	−0.03 (0.877)	—	0.38 (0.009)*	0.58 (<0.001)*	0.039†
INFERIOR	SUPERIOR MS (dB), Linear	0.24 (0.256)	0.03 (0.903)	—	0.04 (0.867)	0.07 (0.738)	—	0.56 (<0.001)*	0.55 (<0.001)*	0.906
	SUPERIOR MS (dB), Log	0.23 (0.265)	0.02 (0.924)	—	0.03 (0.904)	0.06 (0.768)	—	0.57 (<0.001)*	0.58 (<0.001)*	0.841
	SUPERIOR MS (I/L), Linear	0.27 (0.200)	0.10 (0.638)	—	0.01 (0.949)	0.03 (0.890)	—	0.44 (0.003)*	0.49 (<0.001)*	0.294
	SUPERIOR MS (I/L), Log	0.27 (0.195)	0.10 (0.651)	—	0.01 (0.956)	0.03 (0.890)	—	0.45 (0.002)*	0.51 (<0.001)*	0.247

*   P < 0.05, linear and logarithmic regression analyses in each of normal, glaucoma suspect, glaucoma groups.

†   P < 0.05, Hotelling's t-test in glaucoma group.

Figure 2.

 

Scatterplots showing the relationship between TD- and SD-OCT RNFL measurements and corresponding MS expressed in decibels and 1/L by linear and logarithmic regression analyses in glaucomatous eyes. (A–D) GLOBAL RNFL versus GLOBAL MS, (E–H) SUPERIOR RNFL versus INFERIOR MS, and (I–L) INFERIOR RNFL versus SUPERIOR MS according to TD- and SD-OCT measurements expressed in decibels and 1/L. Pearson correlation coefficients for each OCT were compared by using Hotelling's t-test. *P < 0.05, linear and logarithmic regression analyses. †P < 0.05, by Hotelling's t-test. Solid lines and black circles: SD-OCT; dashed lines and white circles: TD-OCT.

Discussion

Anatomic structure represented by RNFL thickness, and the functional VF, are altered at different times in glaucoma development. Structural changes or RNFL defects precede functional visual loss early in glaucoma progression, and RNFL defects can be detected by red-free RNFL photography before the onset of VF problems. ^15,16 One study showed that at least 25% to 35% of retinal ganglion cells died before visual dysfunction was noted in patients with glaucoma. ¹⁷ However, some reports have suggested that a functional deficit or structural loss can be the first finding in the early stages of glaucoma. ^18,19 Furthermore, it has been suggested that a combination of functional and structural tests can improve the sensitivity of glaucoma detection. ^20,21 Therefore, it is of critical importance to know the degree, both quantitatively and objectively, by which damage to any specific structure affects specific VF parameters and how far glaucomatous damage has progressed, to aid in an early diagnosis and to prevent irreversible glaucomatous damage. 

In other studies, various models, including linear or nonlinear regression analyses of glaucoma progression have been used to investigate the relationship between structural parameters expressed on a linear scale and VF measurements, such as retinal sensitivity (decibels or 1/L) or MD (decibels or l/L). ^{19,22 –32} The results have shown that the relationship between structure and function was linear when a functional loss was represented on a linear (1/L) scale, but curvilinear when a logarithmic (decibel) scale was used. ^{19,22 –27} We analyzed both linear and logarithmic regression models when relating RNFL (expressed on a linear scale) to MS values, represented on either a linear or a logarithmic scale, in the normal, glaucoma suspect, and glaucoma groups. One important result was that, in glaucomatous eyes, there was no difference in best fit in the regression models, even when significant structure–function relationships were observed in all the GLOBAL, SUPERIOR, and INFERIOR areas. Leung et al. ³⁰ evaluated the association between average RNFL thickness and MD expressed on a decibel scale in all subjects, including normal and glaucomatous eyes, and showed that the correlation between structure and function was similar according to either TD- or SD-OCT. However, it would be more appropriate to classify subjects into subgroups and to analyze these subgroups separately. Therefore, it is necessary to understand more precisely the dependence of any structure–function relationship on particular perimetric parameters, a scale appropriate to each parameter, the best type of measurement device, and characteristics of study groups. 

In our present efforts to search for useful structure–function relationships with TD- and SD-OCT, we found that RNFL measurements obtained with either OCT modality were significantly associated with VF sensitivity in glaucomatous eyes. In addition, we observed that the correlation between SUPERIOR RNFL defects and an INFERIOR MS deficit was better with SD-OCT than with TD-OCT, whereas the correlation between INFERIOR RNFL defects and a SUPERIOR MS deficit obtained with SD-OCT was not significantly different from that obtained with TD-OCT. Inferior RNFL defects in glaucoma progression generally precede RNFL problems in other regions. Table 1 shows that both MS values, particularly the SUPERIOR MS in the glaucoma group (SUPERIOR MS, 22.5 dB; INFERIOR MS, 25.8 dB) were lower than in normal eyes (SUPERIOR MS, 25.7 dB; INFERIOR MS, 27.8 dB). It is thus possible that those in the glaucoma group were in mild-to-moderate stages of disease. Therefore, a possible explanation for the regional discrepancy in correlation is that SD-OCT can detect mild-to-moderate glaucoma better than TD-OCT. 

In other words, TD-OCT can misdiagnose patients with mild-to-moderate glaucoma as having severe glaucoma. In clock-hour segment 3, a part of the nasal quadrant, we found that RNFL measurement with TD-OCT was not greater than that obtained by SD-OCT (TD-OCT, 57.0 μm; SD-OCT, 57.8 μm). This was not in accordance with comparative RNFL data in other clock-hour segments. It is possible that the variability of RNFL thickness in clock-hour segment 3 recorded by TD-OCT results in an error in the real value of SUPERIOR RNFL thickness. As previous reports have noted, RNFL thickness measurements in the nasal quadrant appear to be more variable than in other quadrants, ^6,7 especially in clock-hour segment 3 of glaucomatous eyes. ⁷ As Knighton and Qian ³³ have suggested, it is possible that the light beam used for RNFL assessment is dimmer in the nasal quadrant than in other areas, owing to the angle of incidence of the beam, and so delineation of RNFL thickness is reduced in accuracy and reliability. Also, the slope of the retinal surface at the nasal side of the optic nerve head may be greater than at the other side, ⁷ leading to a higher level in measurement variation compared with that of other parts. 

Although RNFL measurements obtained with TD-OCT correlated strongly with those of SD-OCT, we found that RNFL assessments measured by SD- and TD-OCT were rather discrepant. The RNFL thickness measured by TD-OCT was greater than that assessed by SD-OCT in both individual segments and the four quadrants and for the average of all clock-hour segments (with the exception of a few individual segments) in all three groups. This finding is generally consistent with the results of other studies. ^12,14,30,34 Possible explanations for this phenomenon are outlined below. RNFL thickness measurements reflect the presence of two principal components, which are, first, the axons of the retinal ganglion cells, and, second, all other structures including astrocytes, the processes of Müller cells, and blood vessels. Thus, we postulate that blood vessels tend to be distributed above the RNFL rather than within the layer, and connective tissues derived from astrocytes and Müller cells would lie between the retinal ganglion cell layer and the RNFL. It has been reported that the segmentation algorithms used to measure RNFL thickness differ between the two OCT modalities, in that SD-OCT identifies the bottom of the nerve fiber layer, whereas TD-OCT localizes the top of the ganglion cell layer. ³⁴ These assumptions and differences in the segmentation algorithms of OCT instruments may explain why SD-OCT provides measurements reflecting a more accurate delineation of the pure RNFL margin. Thus, RNFL thickness measurements are reduced when obtained with SD-OCT compared with TD-OCT. 

There are several limitations to our work. Our study included only a small number of eyes in each group, and further work with more eyes is needed. Owing to our small sample size, the suggested ability of SD-OCT to detect structural changes in mild-to-moderate stages of glaucoma must be confirmed in larger clinical trials. We investigated the structure–function relationship between broad RNFL sectors and VF regions with both linear and logarithmic regression analyses. Thus, we have commenced the creation of a new and detailed map between VF points and sectoral RNFL thicknesses, for different stages of glaucoma's progression, with second- and third-order models as well as linear and logarithmic regression analyses based on SD-OCT RNFL measurements. 

In conclusion, our results indicate that SD-OCT may allow a more accurate assessment of structure–function relationships, compared with TD-OCT, when used to measure inferior retinal sensitivity and superior RNFL defects in glaucomatous eyes. SD-OCT may thus be valuable to investigate structure–function relationships in detail. The high sensitivity of SD-OCT may indicate the utility of the instrument as a diagnostic tool for early detection of glaucoma. 

Footnotes

 Presented at the 101th annual meeting of the Korean Ophthalmological Society, Ilsan, Gyeonggi-do, Korea, April 2009.

Footnotes

 Disclosure: J.R. Lee, None; J.W. Jeoung, None; J. Choi, None; J.Y. Choi, None; K.H. Park, None; Y. Kim, None