Analysis of Neuroretinal Rim by Age, Race, and Gender Using High-Density Three-Dimensional Spectral Domain Optical Coherence Tomography

Abstract

Purpose

To evaluate the relationship between age, race, and gender with the neuroretinal rim using high-density spectral domain optical coherence tomography (SD-OCT) optic nerve volume scans of normal eyes.

Methods

256 normal subjects underwent Spectralis (Heidelberg Engineering, Heidelberg, Germany) SD-OCT optic nerve head volume scans. One eye was randomly selected and analyzed for each subject. Using custom-designed software, the neuroretinal rim minimum distance band (MDB) thickness was calculated from volume scans, and global and quadrant neuroretinal rim thickness values were determined. The MDB is a three-dimensional neuroretinal rim band comprised of the shortest distance between the internal limiting membrane (ILM) and the termination of the retinal pigment epithelium/Bruch’s membrane (RPE/BM) complex. Multiple linear regression analysis was performed to determine the associations of age, race, and gender with neuroretinal rim MDB measurements.

Results

The population was 57% female and 69% Caucasian with a mean age of 58.4 ±15.3 years. Mean MDB thickness in the normal population was 278.4 ± 47.5 μm. For this normal population, MDB thickness decreased by 0.84 μm annually (p <0.001). African Americans had thinner MDBs compared to Caucasians (p= 0.003). Males and females had similar MDB thickness values (p= 0.349).

Conclusions

Neuroretinal rim MDB thickness measurements decreased significantly with age. African Americans had thinner MDB neuroretinal rims than Caucasians.

Précis

Neuroretinal rim minimum distance band thickness is significantly lower in older subjects and African Americans compared to Caucasians. It is similar in both genders.

Spectral domain optical coherence tomography (SD-OCT) has become an integral part of the clinical evaluation for glaucoma because it can objectively measure the neuroretinal rim, the ganglion cell region, and the retinal nerve fiber layer (RNFL), all of which are known to decrease with glaucoma.^1–12 Since OCT imaging can show structural changes years before functional visual field (VF) loss, structural tests remain a cornerstone in the evaluation of glaucoma patients.^13–16 However, clinicians who use neuroretinal rim structural tests need to be aware of how much nerve tissue loss is expected for normal aging and how much racial variation exists in SD-OCT parameters in order to distinguish normal aging changes and racial variation from glaucomatous changes.

Most SD-OCT studies which look at the effect of age, race, and gender on structural parameters have focused on RNFL thickness.^{10–12,17–23} These studies have found that increasing age is associated with RNFL thickness decreases between 0.18 and 0.44 μm annually in normal adult patients.^{10–12,17–21} Multiple studies reported that gender is not significantly correlated with RNFL thickness,^19,22,23 while one reported that women had thicker RNFL than men.²⁴ Two studies reported that Caucasians had thinner RNFL thickness measurements than other racial groups.^18,19 However, RNFL thickness measurements through SD-OCT have a high rate of imaging artifacts, and up to 46.3% of RNFL thickness scans have artifacts, which can be caused by decentration errors, posterior vitreous detachments, and epiretinal membranes.^25,26

In contrast, SD-OCT studies which look at the effect of age, race, and gender on the neuroretinal rim are few and are limited to two studies which used a low-density scan protocol and evaluated the Bruch’s Membrane Opening Minimum Rim Width (BMO-MRW) parameter, which has not been consistently shown to be diagnostically better than the RNFL thickness parameter.^22,23,27 To our knowledge, there are no SD-OCT studies which use a high-density scan protocol to evaluate the effects of age, race, and gender on the neuroretinal rim. Therefore, this current study utilizes the minimum distance band (MDB) neuroretinal parameter, which is derived from high-density optic nerve scans and which has been shown to be diagnostically better than RNFL thickness, especially in the nasal, temporal, and superonasal regions.^29–31 Reproducibility of MDB thickness is high, with an intertest variability of 0.84%.²⁹ Like other 3D neuroretinal rim parameters, the MDB thickness uses SD-OCT imaging to determine the borders of the neuroretinal rim. These other 3D parameters include the minimum circumpapillary band (MCB), which defines the neuroretinal rim as the distance between the internal limiting membrane (ILM) and the retinal pigment epithelium (RPE),³² and the BMO-MRW, which defines the neuroretinal rim as the distance between the ILM and the BMO.^22,23,27 However, although Bruch’s membrane may be visible on SD-OCT, it is often indistinguishable from the RPE as noted by an international panel of SDOCT experts.³³ In addition, the average thickness of the BM is 1–5 μm, whereas the axial resolution of the Spectralis SD OCT machine is 7 μm, making it hard to reliably discern BMO in SD-OCT images. Therefore, the termination of the RPE/BM complex may be a more reliable descriptor of the neuroretinal rim border seen in SD-OCT imaging. The MDB neuroretinal rim parameter is thus defined as the shortest distance between the ILM and the termination of the RPE/Bruch’s membrane (RPE/BM) complex (Figure 1).^28–31 Another fundamental difference between the MDB and the BMO-MRW lies in the image acquisition protocol. The MDB thickness is calculated from high-density raster scans comprised of 193 B-scans resulting in 100 calculated points along the disc border, while the BMO-MRW uses lower-density radial scans comprised of 24 B-scans, resulting in 48 calculated points along the disc border. Additionally, the high-density raster scan protocol allows for the generation of multiple measurements from a single scan, including rim thickness, rim area, and rim volume.^28–31 In contrast to the macular region where only 50% of retinal ganglion cells (RGC) reside, 100% of RGC axons must pass through the neuroretinal rim and they account for almost all the tissue in this MDB region (i.e. 94% nerve axons and 5% astrocytes).³⁴ Since the MDB measures the minimum space through which this trajectory of nerve axons must travel to reach the brain, the MDB is a good surrogate measurement of neuroretinal rim tissue^28–30 and affords a good model to assess to effects of age, race, and gender on optic nerve tissue.

Figure 1.

The neuroretinal rim minimum distance band (MDB) high density raster scan protocol compared to the Bruch’s Membrane Opening-Minimum Rim Width (BMO-MRW) radial scan protocol. (A): MDB thickness scan of a glaucomatous eye not included in this study. Each of the horizontal green lines represents one of the 193 raster scans (top left), which are then used to reconstruct the neuroretinal rim through 100 points (red dots, top center), defined as the termination of the retinal pigment epithelium/Bruch’s membrane (RPE/BM) complex. The last image (top right) shows one of the 193 B-scans, with the termination of the retinal pigment epithelium/Bruch’s membrane (RPE/BM) complex delineated in red, and the internal limiting membrane (ILM) in green. The shortest distance between the ILM and the RPE/BM complex is represented by the yellow arrow, although the final average MDB thickness is calculated from a 100-point 3D reconstruction of the ILM and RPE/BM complex terminations. (B): BMO-MRW scan of a healthy eye. Each green line represents one of the 24 radial scans used to reconstruct the disc margin at 48 points (red dots, bottom left), which is defined as the border of the BMO. A cross-section at one of those points is shown (bottom right), with the ILM marked by a red line, and the BMO by a red dot, while a blue arrow delineates the BMO-MRW.

To the best of our knowledge, the effects of age, race, and gender on the high-density neuroretinal MDB parameter have not yet been reported. The aim of this study was to determine the relationships between these variables and MDB thickness in a multi-ethnic population using 3D high-density volume scans acquired by the Spectralis SD-OCT machine (Heidelberg Engineering, Heidelberg, Germany).

Methods

Participants and Associated Testing

All subjects were recruited from the Glaucoma Service at the Massachusetts Eye and Ear Infirmary between April 2009 and January 2016, though all were deemed to have no glaucomatous damage This prospective study protocol was approved by the Massachusetts Eye and Ear Infirmary Institutional Review Board, and informed consent and Health Insurance Portability and Accountability Act (HIPAA) forms were signed by all study participants. This study is a cross-sectional sampling of this prospective study. All subjects underwent a complete eye examination by a glaucoma specialist (T.C.C.), which included history, best corrected visual acuity testing, Goldmann applanation tonometry, slit-lamp biomicroscopy, gonioscopy, ultrasonic pachymetry (PachPen, Accutome Ultrasound Inc., Malvern, PA), dilated ophthalmoscopy, stereo disc photography (Visucam Pro NM; Carl Zeiss Meditec, Inc), visual field (VF) testing (Swedish Interactive Threshold Algorithm 24–2 test of the Humphrey visual field analyzer 750i; Carl Zeiss Meditec, Inc), as well as RNFL thickness scans (HRA/Spectralis software version 5.4.8.0 Heidelberg Engineering, GmbH, Heidelberg, Germany).

Patients were only included if their VF tests were reliable: < 33% fixation losses, < 20% false positives, and < 20% false negatives; if their RNFL scans had a clear fundus image with good optic disc and scan circle visibility prior to and during image acquisition, RNFL visible and without interruptions, and a continuous scan pattern without missing or blank areas. All included study patients had normal eyes except for mild cataracts, had best-corrected visual acuities of 20/40 or better, had spherical equivalent refractions within ± 5.0 D, had intraocular pressures (IOP) <21 mmHg, and had normal VF testing with normal Glaucoma Hemifield Tests. Normal visual field tests did not have a cluster of three −5 dB or more abnormal spots on the same side of the horizontal meridian and did not have a cluster of −5 dB and −10 dB or more abnormal spots on the same side of the horizontal meridian on the pattern standard deviation map. Although all patients had normal optic nerves, these normal patients were divided into two subgroups based on their cup-to-disc ratios (CDR), which were determined by subjective assessments by a glaucoma specialist (T.C.C.): Subgroup A with normal nerves and Subgroup B with normal disc variations (i.e. physiologic cupping). Physiologic cupping was defined as having CDR > 0.4 for Caucasians and Asians and > 0.6 for African Americans and Hispanics, with normal VF test results and IOP <21 mmHg. Patients were excluded if they had a CDR asymmetry > 0.2, or if scans had a manufacturer signal strength ≤ 15 dB. When two eyes of the same patient were eligible for the study, one eye was randomly selected by the investigator using a random number generator.

Spectralis SD-OCT Optic Nerve Volume Scan and MDB Calculations

SD-OCT optic nerve volume scans were performed after pupillary dilation using the Spectralis OCT machine, which relies on an 870-nm superluminescent diode source. The Spectralis OCT’s automatic real-time (ART) function and eye tracking system was used to increase image quality. The high-density 3D optic nerve head volume scan protocol consists of 193 B-scans in a raster pattern over an area 20 degrees by 20 degrees (approximately 6 mm x 6 mm, depending on the patient’s refraction) centered on the optic nerve head. To analyze this volumetric dataset, a custom-designed software was written, using C++ with OpenCV, ITK, and VTK libraries (ET). The algorithm automatically segments B-scans to reconstruct the ILM and RPE/BM in 3D. Segmentation errors were identified manually and automatically interpolated by the software. The neuroretinal rim was determined from the termination of the RPE/BM complex at 100 circumferential points, and the MDB thickness was calculated by measuring the closest distances from these points to the ILM (Figure 1). All calculations were performed in real space, though the images in Figure 1 are stretched by a factor of three for display purposes. Two adjacent points on the neuroretinal rim and their closest points on the ILM formed pairs of triangles. The MDB area was calculated by measuring the area of the triangle pairs around the rim. Global, quadrant, and octant averages of the thickness and area measurements were determined via the arithmetic mean. Further details of this MDB program have been described in previous studies.^29,30

Calculation of Percentage of MDB Thinning Per Year

We calculated percentage of nerve tissue loss per year for MDB thickness in order to better compare our results to those of other neuroretinal rim or RNFL thickness parameters, whose normal thickness values may have different magnitudes. The annual percentage of MDB loss was defined as annual MDB thinning (i.e. numerator) divided by normal mean MDB thickness measurements (i.e. denominator).

Statistical Analysis

A Chi-square test was used to test whether the race and gender distributions of the two subgroups were different. Simple linear regression analysis was performed to determine the association between age and MDB thickness. A t-test was used to determine the effect of gender on MDB thickness, and to compare the MDB thickness of the normal Subgroup A with that of Subgroup B with normal disc variations. An F-test was performed to determine whether the MDB thickness was different among different races. A Benjamini-Hochberg false discovery rate correction was performed to account for multiple testing of the univariate analyses among all Subgroup A and Subgroup B participants. Furthermore, a multivariate linear regression analysis of MDB thickness was performed, adjusting for age, race, and gender. P-values of < 0.05 were considered statistically significant. Results are shown as the mean ± standard deviation (SD) unless otherwise stated. Finally, an ANOVA F-test was used to determine whether age-related MDB thickness changes were different across the races included in this study.

Results

Subject Characteristics

The study included 256 normal patients: 132 patients in Subgroup A with normal discs and 124 patients in Subgroup B with normal disc variations (i.e. physiologic cupping). The mean age for all 256 patients was 58.4 ± 15.3 years (Table 1). The study population was predominantly female (57%) and Caucasian (69%) and included 134 right eyes and 122 left eyes (Table 1).

Table 1.

Demographics of normal study population (n= 256).

	Overall Normal Study Group	Subgroup A: Normal Discs	Subgroup B: Normal Disc Variations	P*
Mean Age ± SD, years	58.4 ± 15.3	56.8 ± 16.3	60.0 ± 14.0	0.093
Number of Subjects: Gender				0.346
Female	145 (57%)	79 (60%)	66 (53%)
Male	111 (43%)	53 (40%)	58 (47%)
Number of Subjects: Race				0.004*
Caucasian	177 (69%)	86 (65%)	91 (73%)	0.197
African-American	31 (12%)	21 (16%)	10 (8%)	0.083
Asian	28 (11%)	11 (8%)	17 (14%)	0.239
Hispanic	17 (7%)	14 (11%)	3 (2%)	0.017*
Other	3 (1%)	0	3 (2%)	N/A
Refractive Error, Diopters	−0.66 ± 1.91	−0.40 ± 1.79	−0.94 ± 2.01	0.025*
Overall	256	132	124

SD= Standard Deviation.

P is the p-value of testing that the proportion of subjects from each demographic is the same in subgroups A and B, measured with a Chi-square test for gender and race, and the p-value of testing that the means of age and refractive error among groups A and B are equal using a t-test.

^*= statistically significant P <0.05

MDB Thickness Measurements

The mean MDB thickness was 278.4 ± 47.5 μm for all subjects (Table 2). Most normal eyes did not follow the ISNT rule, which states that the inferior rim is the thickest, followed by the superior rim, the nasal rim, and then the temporal rim as the thinnest.³⁵ In the overall normal group of 256 patients, only 71 (28%) of 256 eyes obeyed the ISNT rule. Of the normal subgroups, the ISNT rule was only valid for 40 (30%) of 132 eyes of normal Subgroup A and 31 (25%) of 124 eyes of Subgroup B with normal disc variations. In contrast, the mean MDB thickness values, averaged as an entire group, did follow the ISNT rule (Table 2). Global MDB was thicker in Subgroup A with normal discs (302.5 ± 42.2 um) compared to Subgroup B (252.7 ± 38.4 um) and across all quadrants and sectors (p <0.001, Table 3).

Table 2.

Neuroretinal rim minimum distance band (MDB) thickness and area in the normal study population (n= 256) by quadrant and sector.

	MDB Thickness ± SD (95% CI, um)	MDB Area ± SD (95% CI, mm2)
Global	278.4 ± 47.5 (272.6–284.2)	1.824 ± 0.409 (1.774–1.874)
Inferior Quadrant	312.8 ± 60.6 (305.4–320.2)	0.526 ± 0.142 (0.509–0.543)
Superior Quadrant	297.8 ± 62.4 (290.2–305.5)	0.505 ± 0.153 (0.487–0.524)
Nasal Quadrant	281.5 ± 56.9 (274.6–288.5)	0.461 ± 0.144 (0.444–0.479)
Temporal Quadrant	222.4 ± 47.6 (216.6–228.2)	0.332 ± 0.105 (0.319–0.345)
Inferior Nasal Sector	325.8 ± 63.3 (318.0–333.5)	0.278 ± 0.080 (0.268–0.288)
Inferotemporal Sector	299.8 ± 68.9 (291.3–308.2)	0.248 ± 0.086 (0.238–0.259)
Superior Nasal Sector	304.4 ± 66.7 (296.2–312.5)	0.254 ± 0.090 (0.243–0.265)
Superotemporal Sector	298.0 ± 63.9 (290.2–305.9)	0.252 ± 0.086 (0.241–0.262)

MDB= Minimum Distance Band, SD= Standard Deviation, CI= Confidence Interval

Table 3.

Neuroretinal rim minimum distance band (MDB) in the normal Subgroup A (normal discs, n=132) and normal Subgroup B (normal disc variations, n= 124), by quadrant and sector.

	MDB Thickness ± SD (95% CI, μm)
	Subgroup A: Normal Discs (n= 132)	Subgroup B: Normal Disc Variations (n= 124)	P*
Global	302.5 ± 42.2 (295.2–309.7)	252.7 ± 38.4 (246.0–259.5)	<0.001*
Inferior Quadrant	337.2 ± 57.3 (327.4–346.9)	286.9 ± 53.0 (277.5–296.2)	<0.001*
Superior Quadrant	328.7 ± 54.0 (319.5–337.9)	265.0 ± 53.6 (255.5–274.4)	<0.001*
Nasal Quadrant	304.6 ± 53.2 (295.5–313.7)	257.0 ± 50.2 (248.1–265.8)	<0.001*
Temporal Quadrant	240.5 ± 44.8 (232.9–248.2)	203.1 ± 42.8 (195.5–210.6)	<0.001*
Inferior Nasal Sector	348.5 ± 62.4 (337.9–359.2)	301.6 ± 54.9 (291.9–311.2)	<0.001*
Inferotemporal Sector	325.4 ± 61.6 (314.9–335.9)	272.4 ± 65.9 (260.8–284.0)	<0.001*
Superior Nasal Sector	336.2 ± 59.0 (326.1–346.3)	270.5 ± 57.1 (260.4 – 280.5)	<0.001*
Superotemporal Sector	328.2 ± 52.9 (319.2–337.2)	265.9 ± 59.0 (255.5–276.3)	<0.001*

MDB= Minimum Distance Band, SD= Standard Deviation, CI= Confidence Interval

P is the chance that the MDB thickness is equal in Subgroup A and Subgroup B. A false discovery rate correction was applied to all calculations.

^*= Statistically significant P <0.05.

Association of Age and MDB Thickness Measurements

MDB thickness decreased significantly with age for the overall normal population (p= 0.003), at a rate of 0.71 ± 0.19 μm per year (Table 4). The inferior, superior, and nasal quadrants displayed similar results (p <0.011), while temporal MDB thickness did not show a significant correlation with age (p= 0.077, Table 4). Figure 2 shows the decline of MDB thickness with increasing age in the overall normal study population. Results were similar in normal Subgroup A with normal discs, while Subgroup B with normal disc variations did not show a significant decline with age (p= 0.955, Table 4).

Table 4.

Effect of age on neuroretinal rim minimum distance band (MDB) thickness in the normal study population (n=256), the normal Subgroup A (n=132), and the normal Subgroup B with normal disc variations (n=124), by quadrant and sector.

	Overall Normal Study Group (n= 256)		Subgroup A: Normal Discs (n= 132)		Subgroup B: Normal Disc Variations (n= 124)
	MDB Thinning Per Year ± SD (μm)	P*	MDB Thinning Per Year (μm)	P*	MDB Thinning Per Year (μm)	P*
Global	0.71 ± 0.19	0.003*	0.81 ± 0.22	0.003*	0.14	0.955
Inferior Quadrant	0.81 ± 0.24	0.006*	1.04 ± 0.30	0.003*	0.06	0.955
Superior Quadrant	0.75 ± 0.25	0.011*	0.80 ± 0.28	0.021*	0.15	0.955
Nasal Quadrant	0.87 ± 0.23	0.003*	1.02 ± 027	0.003*	0.25	0.955
Temporal Quadrant	0.41 ± 0.19	0.077	0.39 ± 0.24	0.266	0.12	0.955
Inferior Nasal Sector	0.74 ± 0.26	0.011*	0.93 ± 0.33	0.021*	0.09	0.955
Inferotemporal Sector	0.87 ± 0.28	0.003*	1.15 ± 0.32	0.027*	0.03	0.955
Superior Nasal Sector	0.97 ± 0.27	0.008*	1.13 ± 0.30	0.025*	0.21	0.955
Superotemporal Sector	0.52 ± 0.26	0.088	0.51 ± 0.28	0.211	0.02	0.955

MDB= Minimum Distance Band

P is the chance that the change due to age is not statistically different from a slope of 0, indicating no change due to age. A false discovery rate correction was applied to all calculations.

^*= Statistically significant P <0.05.

Calculations were performed using a univariate analysis model.

Figure 2.

Scatter-plot showing the relationship between age (years) and the total mean neuroretinal rim minimum distance band (MDB) thickness (μm) in the normal study population (n= 256). Increasing age was significantly associated with decreasing MDB thickness (p < 0.001). Subgroup A with normal discs is shown in blue, Subgroup B with normal disc variations is shown in red.

Association of Gender and MDB Thickness Measurements

There was no difference in MDB thickness between males and females in the overall study population or in either subgroup, globally or in any quadrant or sector (p> 0.162, Table 5).

Table 5:

Effect of gender on neuroretinal rim minimum distance band (MDB) thickness in the normal study population (n=256), by quadrant and sector.

	MDB Thickness ± SD (um)
	Females	Males	P*
Global	279.5 ± 48.2	276.9 ± 46.7	0.790
Inferior Quadrant	313.9 ± 63.6	311.3 ± 56.7	0.819
Superior Quadrant	302.1 ± 62.5	292.2 ± 62.2	0.299
Nasal Quadrant	281.0 ± 59.0	282.2 ± 54.3	0.897
Temporal Quadrant	222.2 ± 46.0	222.6 ± 49.9	0.946
Inferior Nasal Sector	326.7 ± 64.2	324.6 ± 62.3	0.856
Inferotemporal Sector	301.4 ± 74.4	297.7 ± 61.6	0.790
Superior Nasal Sector	308.9 ± 66.1	298.4 ± 67.2	0.299
Superotemporal Sector	303.8 ± 63.0	290.6 ± 64.7	0.162

MDB= Minimum Distance Band

P is the chance that the difference between mean MDB measurements of males and females is not statistically different from a slope of 0, indicating no difference. A false discovery rate correction was applied to all calculations.

^*= Statistically significant P <0.05.

Calculations were performed using a univariate analysis model.

Association of Race and MDB Thickness Measurements

Mean MDB thickness was highest among Hispanics (286.2 ± 57.7 μm), followed by Caucasians (283.7 ± 44.0 μm), Asians (269.5 ± 40.1 μm), and African Americans (258.6 ± 57.5 μm) (p= 0.011, Table 6). African Americans had the thinnest global, temporal, and inferior nasal MDB, while Asians had the thinnest nasal MDB, and Hispanics had the thickest MDB in all four regions. P-values for testing of mean equality of MDB measurements across races were p= 0.011, 0.008, 0.026, 0.006 for global, temporal, inferior nasal and nasal sections. MDB thickness in other quadrants and sectors was similar between all four races (p> 0.051, Table 6).

Table 6:

Effect of race on neuroretinal rim minimum distance band (MDB) thickness in the normal study population (n=256), by quadrant and sector.

	MDB Neuroretinal Rim Thickness ± SD (um)
	Caucasian	African American	Asian	Hispanic	P*
Number of Subjects	177	31	28	17
Global	283.7 ± 44.0	258.6 ± 57.5	269.5 ± 40.1	286.2 ± 57.7	0.011*
Inferior Quadrant	317.7 ± 58.3	293.7 ± 67.4	304.8 ± 50.7	320.2 ± 77.3	0.144
Superior Quadrant	300.3 ± 59.2	283.5 ± 75.1	295.1 ± 60.0	316.2 ± 69.9	0.144
Nasal Quadrant	289.7 ± 53.3	264.9 ± 62.2	258.9 ± 58.0	280.0 ± 60.1	0.006*
Temporal Quadrant	227.8 ± 44.2	193.4 ± 52.9	220.9 ± 48.4	229.0 ± 52.3	0.008*
Inferior Nasal Sector	332.7 ± 59.7	304.3 ± 68.8	312.2 ± 49.2	330.6 ± 90.2	0.026*
Inferotemporal Sector	302.8 ± 68.5	282.8 ± 74.2	309.3 ± 68.0	309.3 ± 68.0	0.615
Superior Nasal Sector	306.0 ± 61.7	296.1 ± 79.9	303.6 ± 74.8	317.9 ± 77.0	0.342
Superotemporal Sector	301.8 ± 60.2	276.0 ± 78.5	295.5 ± 60.6	318.0 ± 68.0	0.051

MDB= Minimum Distance Band

SD= Standard deviation.

Results are expressed as mean ± standard deviation.

P is the chance that the MDB thickness is the same across all races, measured with an F-test using ANOVA. A false discovery rate correction was applied to all calculations.

^*= Statistically significant P <0.05.

Calculations performed using a univariate analysis model.

Changes in MDB Thickness Adjusted for Age, Race, and Gender

A multivariate analysis was performed on MDB thickness in the overall study group (n= 256), adjusted for age, race using Caucasians as a reference group, and gender using females as a reference group. The MDB thickness decreased globally (0.84 ± 0.19 um per year, p <0.001), and across all quadrants and sectors with age (p <0.014, Table 7) at a higher rate after adjusting for race and gender. African Americans had thinner MDB than Caucasians globally (p= 0.003) and across the inferior, nasal, and temporal quadrants, and the inferonasal and superior temporal sectors (p= 0.003, 0.028, 0.010, <0.001, 0.013, and 0.043, respectively), whereas the superior quadrant, superonasal and inferotemporal sectors were similar among African Americans and Caucasians (p = 0.095, 0.434, and 0.111, respectively). Asians also had significantly thinner MDB compared to Caucasians in the global, nasal, and inferonasal measurements (p= 0.031, <0.001, and 0.032, respectively). Hispanics had similar MDB thickness to Caucasians globally (p= 0.826) and across all quadrants and sectors (p > 0.225). After adjusting for age and race, males and females had similar MDB thickness globally and in all quadrants and sectors (p > 0.050) except the superotemporal sector, where males had thinner MDB (p= 0.045). Age-related MDB thinning was not significantly different across all races (p > 0.235).

Table 7.

Effect of age on neuroretinal rim minimum distance band (MDB) thickness adjusted for race and gender in the overall study population (n= 256), with annual rate of (MDB) decline in the normal study population (n= 256), by quadrant and sector.

	MDB Thinning Per Year ± SD (μm)	P*	MDB Thinning Per Year (%)
Global	0.84 ± 0.19	<0.001*	0.301
Inferior Quadrant	0.94 ± 0.25	<0.001*	0.300
Superior Quadrant	0.88 ± 0.26	0.001*	0.291
Nasal Quadrant	1.07 ± 0.22	<0.001*	0.381
Temporal Quadrant	0.48 ± 0.19	0.014*	0.216
Inferior Nasal Sector	0.92 ± 0.26	<0.001*	0.282
Inferotemporal Sector	0.96 ± 0.29	0.001*	0.319
Superior Nasal Sector	1.11 ± 0.27	<0.001*	0.359
Superotemporal Sector	0.67 ± 0.26	0.012*	0.221

MDB= Minimum Distance Band

P is the chance that the change due to age is not statistically different from a slope of 0, indicating no change due to age. A false discovery rate correction was applied to all calculations.

^*= Statistically significant P <0.05.

Calculations were performed using a multivariate analysis model, with MDB thickness as the dependent variable, and age, race, and gender as the independent variables.

MDB thinning ratio is calculated by dividing the annual rate of MDB thinning (column 1) by the mean MDB thickness in the overall population (Table 2).

MDB Area Measurements

The mean MDB area in the study population was 1.824 ± 0.409 mm² and overall average group values for MDB area followed the ISNT rule (Table 2). In the normal Subgroup A, the mean MDB area was 1.975 ± 0.410 mm². The superior and inferior quadrants were similar in size (0.564 ± 0.148 and 0.562 ± 0.137 mm², respectively), followed by the nasal (0.493 ± 0.144 mm²) and temporal quadrants (0.357 ± 0.116 mm²). In Subgroup B with normal disc variations, the global MDB area was 1.664 ± 0.342 mm². The average subgroup quadrant values followed the ISNT rule (0.487 ± 0.137, 0.443 ± 0.132, 0.428 ± 0.137, and 0.305 ± 0.084 mm², respectively).

Discussion

To our knowledge, this is the first study that describes the relationship of age, race, and gender with the 3D SD-OCT MDB neuroretinal rim parameter. On average, global MDB thickness decreases 0.84 ± 0.19 μm per year (Table 7), with similar rates between men and women (p> 0.162, Table 5). In terms of ethnic differences, African Americans had significantly thinner MDB values compared to Caucasians (258.6 ± 57.5 μm versus 283.7 ± 44.0 μm, p= 0.003). Although differences were not significant, Hispanics had larger global MDB thickness values (286.2 ± 57.7 μm) and Asians had thinner MDB values (269.5 ± 40.1) compared to Caucasians (Table 6). Since neuroretinal rim thickness measurements such as the MDB thickness and BMO-MRW may be considered diagnostically superior to area measurements, such as MDB area or BMO area, this paper’s discussion focuses on the MDB thickness.^31,36,37

Like RNFL thickness measurements, MDB neuroretinal rim thickness also decreases with age (Tables 4, 7,8). Age-related RNFL thinning has been reported to be 0.18 – 0.44 μm per year as measured by SD-OCT.^{10–12,17–20,38} Since RNFL and MDB measure different anatomical regions and therefore have different normal mean values, comparing rates of percentage decline instead of absolute value decline would make comparison of these two parameters easier. Therefore, past cross-sectional studies have reported that annual RNFL thinning ranges between 0.15% by Alasil et al. to 0.38% by Celebi et al.^{12,17–20,38} Vianna et al. reported a decline of 0.46% per year in 37 normal adults over the course of a 4-year (range 2–6 years) longitudinal study.¹¹ In order to compare MDB thickness to RNFL thickness and BMO-MRW, we converted the annual decline measured in microns into a proportion and presented it in Table 8. The MDB thickness decreased by an average annual rate of 0.25% in our 256 normal study subjects, indicating that our results on the MDB age-related thinning are in line with the existing literature on RNFL age-related thinning (Table 8).

Table 8.

Annual rate of neuroretinal rim minimum distance band (MDB) decline in the normal study population (n= 256) for thickness, by quadrant and sector.

	MDB Thinning Per Year (%)
	Overall Normal Study Group (n= 256)	Subgroup A: Normal Discs (n= 132)	Subgroup B: Normal Disc Variations (n= 124)
Global	0.253	0.269	0.056
Inferior Quadrant	0.258	0.309	0.019
Superior Quadrant	0.251	0.243	0.057
Nasal Quadrant	0.308	0.337	0.097
Temporal Quadrant	0.183	0.163	0.057
Inferior Nasal Sector	0.228	0.267	0.030
Inferotemporal Sector	0.289	0.353	0.010
Superior Nasal Sector	0.319	0.336	0.076
Superotemporal Sector	0.176	0.156	0.008

MDB= Minimum Distance Band.

Annual MDB Thinning ratio is calculated by dividing the average thinning with age (Table 4), by the mean MDB thickness (Tables 2 and 3), and multiplying by 100.

Rates for age-related decline of the BMO-MRW neuroretinal rim parameter have been reported at 1.34 to 1.92 μm per year, similar to the MDB thinning of 0.84 μm per year reported in this study (Table 7).^11,12,19 When using the same methodology as the current study to calculate rates of percentage decline, BMO-MRW studies reported a decline of 0.40 to 0.63% per year, which is higher but similar to the 0.25% decline reported in this study for global MDB thickness (Table 8).^11,12,19 In a confocal scanning laser tomography (CSLT) study, Enders et al. more recently reported a decline of 0.80 μm per year (or 0.34% per year) in adults with a large optic nerve head, defined as having an area ≥ 2.45 mm².³⁹ This rate of 0.34% per year by CSLT is similar to the rate of 0.30% per year by SD-OCT in this study (Tables 7, 8). One reason for the slight difference between percentage decline for the BMO-MRW parameter (0.40 to 0.63% per year)^11,12,19 and the MDB thickness parameter (0.30% per year, Tables 7 and 8) may be that the BMO-MRW and MDB thickness measurements are procured differently. For example, the BMO-MRW low-density scan protocol consists of 24 radial scans, with 25 averages each, while the MDB parameter is derived from a high-density 3D volume scan with 193 raster lines, with 3 averages each.^29,30 The definition of the disc border also differs between the BMO-MRW and the MDB thickness parameter, because the BMO is used for the BMO-MRW parameter and the termination of the RPE/BM complex is used for the MDB parameter. These data acquisition differences may have accounted for the slight difference between BMO-MRW and MDB rates of age-related decline. Nevertheless, this study and the past literature overall seems to suggest that neuroretinal rim parameters by CSLT and SD-OCT appear to have similar percentage rates of decline.

Rates of age-related neuroretinal thinning as measured by SD-OCT in this study are similar to those reported in past histologic studies, which have found a significant age-related decline in the number of RGC axons.^40–43 The estimated mean nerve fiber count is around 0.97 −1.24 million fibers,^41–43 with mean loss of around 4,000 to 5,400 fibers (0.32–0.54%) per year.^40,42,43 The annual rates of thinning measured in this study, 0.253% for MDB thickness and 0.279% for MDB area (Table 8), are similar to those reported in histologic studies, which confirms the good correlation between neuroretinal MDB OCT measurements and histologic nerve fiber counts. The subtle differences between the predicted decay and the observed decay may be due to the presence of non-axonal tissue or even blood vessel artifacts which may influence MDB calculations.²⁸ Although future studies are needed to verify this hypothesis, MDB thickness measurements may better reflect nerve tissue loss compared to RNFL thickness measurements, because the MDB thickness measurements have a higher component of nerve to non-neuronal tissue. For example, primate histology studies indicate the MDB may be comprised of up to 94% nerve axons and only 5% astrocytes,³⁴ while the RNFL is composed of at least 18% glial cells, including Muller cells and astrocytes.⁴⁴ Additionally, while SD-OCT studies suggest that the RNFL thickness measurements may be comprised of almost 48.8% to 65.1% of non-neuronal tissue (i.e. glial cells and blood vessels).⁴⁵ Even though SD-OCT RNFL thickness studies have well substantiated a “floor effect” ranging from 49.2 to 64.7 μm due to glial cells and blood vessels,^29,45 future studies of MDB thickness are needed to verify that the “floor effect” for MDB measurements are indeed lower than that for RNFL thickness measurements. These future studies would further substantiate whether OCT is an accurate form of in vivo histology or not.⁴⁶

Table 4 shows that rates of age-related decline in MDB thickness were similar for all quadrants and sectors except for the temporal quadrant, which did not decline with age (p= 0.077). This is consistent with studies on RNFL age-related thinning, which showed that the thickness of the mean, superior, inferior, and nasal quadrants decreased with age, while the temporal quadrant did not.^12,20,38 After adjusting for gender and race, the temporal quadrant declined significantly with age (p= 0.014), but it had the slowest rate of age-related decline when analyzed for the entire study population compared to other quadrants (i.e. 0.48 μm compared to 0.88 to 1.07 μm yearly, or 0.18% compared to 0.25–0.31% yearly, Table 7). This is similar to results in the BMO-MRW, where all quadrants declined with age and the temporal quadrant showed the slowest rate of decline.¹² One possible explanation for the slower rate of age-related decline for the temporal quadrant is that it contains the papillomacular bundle, which is composed of thinner axons.⁴² Thus, assuming an equal loss in the number of axons across all quadrants, the temporal quadrant would display the least thinning since it has the thinnest axons to start with. Another theory is that slower axonal loss near the fovea may be an evolutionary protective mechanism to preserve central vision, which is supported by the temporal region of the ONH.³⁸

Subjects with normal disc variation (i.e., physiologic cupping) had thinner MDB values than those with normal discs (p <0.001, Table 3) and showed no significant thinning with age (p= 0.955, Table 4), while those with normal discs showed significant MDB thinning with age across all but the temporal quadrant and the superotemporal sector (p <0.028, Table 4). Subgroup B with physiologic cupping was more myopic than Subgroup A with normal discs (Table 1), which may affect MDB thickness measurements due to increased optic disc tilt or peripapillary atrophy. Subgroup B also had significantly fewer Hispanics than Subgroup A (Table 1), which may have contributed to the thinner MDB measured in Subgroup B compared to Subgroup A, since Hispanics had the thickest MDB measurements overall (Table 6). The difference in age-related MDB decline among the two groups may be due to the fact that subjects with physiologic cupping have a lower mean MDB thickness. The normal variability of subjects within each subgroup may also explain the difference in age-related change, since aging may act differently on certain groups or individuals. Additionally, since the main difference between normal Subgroup B patients and normal Subgroup A patients is the larger CDR of Subgroup B, it is likely that this larger CDR may play a role in their having thinner MDBs and their having a smaller percentage decline in neuroretinal rim thickness per year (Table 8). Individuals with a larger CDR sometimes have a larger disc diameter, which means that despite having a similar number of axons, the neuroretinal rim is expected to be thinner in individuals with a larger CDR. A study by Tatham et al. also concluded that small differences in CDR were inversely correlated with large changes in RGC count which in turn affects neuroretinal rim thickness.⁴⁷

In line with previous studies on RNFL thickness, our study showed that gender did not affect MDB thickness measurements (p = 0.790, Table 5).^{17,19,20,23,48} Like our current study, the literature is also conflicted on the effect of gender on OCT measurements.^49–51 Tun et al. reported a significant relationship between BMO-MRW and gender in a normal Chinese population, with females having thicker measurements.⁴⁹

Table 6 shows that MDB thickness was different among races only in the global, nasal, temporal, and inferior nasal measurements (p <0.027). However, no significant difference in age-related MDB thinning was detected across races (p >0.235), which may be due to a small sample size. A multivariate analysis adjusting for age and gender also showed that African Americans generally have thinner MDB thickness measurements compared to Caucasians globally and across all but the superior quadrant (p <0.029). This is consistent with past studies which have noted thinner temporal RNFL thickness values in African Americans.^18,50,52 However, Knight et al. reported, compared to Caucasians, African Americans had thicker mean and quadrant RNFL measurements in all but the temporal quadrant,⁵⁰ while other studies have reported no significant differences in mean global RNFL thickness between African Americans and Caucasians.^18,20,52 Rhodes et al. found no significant difference in BMO-MRW thickness between subjects of European Descent (ED) and those of African Descent (AD), but reported that the RNFL was thinner in AD subjects in the temporal and superior temporal regions and thicker in the nasal, inferotemporal, inferonasal, and superior nasal regions.⁵² In a longitudinal study of BMO-MRW and RNFL thickness among AD and ED subjects, Bowd et al. found no difference in baseline BMO-MRW, annual BMO-MRW thinning, RNFL thickness, or RNFL thinning in healthy subjects among the two groups, although they did note a faster rate of BMO-MRW thinning in AD “glaucoma suspects,” compared to their ED counterparts.⁵³ Some studies have also found no significant difference in rim area among subjects of different races.^{50, 54}

Although our study found that MDB thickness values in Hispanics were similar to that of Caucasians (p >0.225), it is difficult to say if these results are generalizable, because there were only 17 Hispanic subjects in this study, which makes it difficult to avoid Type II errors. Our study also found that Asians had thinner MDB compared to Caucasians globally, nasally, and inferonasally (p <0.033). Studies on the difference in RNFL thickness between Asians and Caucasians were conflicted.^20,51 Girkin et al. studied RNFL thickness among two Asian ethnicities, Indians and Japanese, and found no difference between mean global RNFL thickness of Japanese and Indian subjects compared to those of European descent or between Japanese and Indian subjects.¹⁸ However, the study found that subjects of Indian descent had thicker RNFL than Europeans across all quadrants, while subjects of Japanese descent had thicker nasal RNFL than those of European descent, but were similar in all other quadrants.²⁰ Another study, by Knight et al., noted that Asians had a thicker RNFL than Europeans across all quadrants except the nasal quadrant, which was similar in thickness among both groups.⁵⁰ One possible explanation for the existence of racial differences in our study is that Hispanics, Asians, and African Americans are believed to have a larger optic disk size compared to Caucasians.⁵⁵ The MDB, which measures neuroretinal rim tissue, may be more affected by disc morphology than the RNFL, leading to differences that are not observed in the RNFL studies. Girkin et al. also hypothesized that the lack of a significant effect of race on the diagnostic performance of SD-OCT may be due to individual differences among subjects of the same race, which may exceed the difference between multiple races.⁵⁴ Another possible explanation may be the small sample size of non-Caucasians in our study (Table 1), which makes finding statistically significant differences more difficult. A post-hoc power analysis revealed that our study does have sufficient power to test whether MDB thickness differences exist among all groups in the global, nasal, temporal, and inferonasal regions (power > 90%). However, when comparing the two largest groups, African Americans and Caucasians, we only had 74% power to detect differences in global thickness, and 97% power to detect differences in the temporal region, with all other regions falling below 65% power. It is important to note that the post-hoc power analysis utilized the mean MDB thickness values in the observed sample (Table 6) as the true value during the calculation, which explains why the power is highest in the regions with the largest difference between African Americans and Caucasians.

Our study has several limitations. As with any cross-sectional study which attempts to evaluate the longitudinal effects of aging, our study results may not accurately reflect the real effects of aging, which is best evaluated in a longitudinal study. Nevertheless, this cross-sectional study still provides useful information, because a longitudinal study over many decades is not possible since SD-OCT has only been commercially available for the past decade or so. Future studies are needed with larger sample sizes of all racial subgroups, in order to better assess if racial differences exist between Caucasians and other groups. No statistically significant difference in age-related decline was detected across races, which can be the result of a small sample size. We performed a power analysis which concluded that with a sample size of 256 participants, equal to the observed sample size, we have 41% power to detect such difference, while a sample size of 580 participants, with race proportions equal to the ones in our observed data, is needed to have 80% power. Additionally, including a larger range of refractive errors would have enabled us to elucidate the effects of myopia or hyperopia on normal MDB thickness and area measurements. Another possible limitation of the study is the use of the default Spectralis pixel conversions when acquiring the images, which may vary with refraction. Therefore, a better study design would have corrected for refractive errors prior to acquiring the scans. Finally, future studies should account for variations in optic disc size, which can affect the size of the RPE/BM termination opening, which in turn may affect MDB neuroretinal rim thickness measurements.

In summary, this study shows that age-related decline in neuroretinal MDB thickness normally occurs at the rate of 0.71 ± 0.19 μm each year (Table 4), which increases to 0.84 ± 0.19 (Table 7) when adjusted for race and gender. Gender does not appear to affect MDB thickness measurements (p >0.162, Table 5). African Americans and Asians had thinner MDB neuroretinal rims compared to Caucasians (p= 0.003 and 0.031, respectively), while MDB thickness measurements for Hispanics were statistically similar to those of Caucasians (p= 0.826). We believe that the results of this study can better inform clinicians on how to account for the effect of normal aging when analyzing MDB thickness measurements over years.