Peripapillary Retinal Nerve Fiber Layer Thickness Determined by Spectral-Domain Optical Coherence Tomography in Ophthalmologically Normal Eyes

Objectives  To evaluate the peripapillary distribution of retinal nerve fiber layer thickness (RNFLT) in normal eyes using spectral-domain optical coherence tomography and to study potentially related factors.

Methods  In 7 institutes in Japan, RNFLT in 7 concentric peripapillary circles with diameters ranging from 2.2 to 4.0 mm were measured using spectral-domain optical coherence tomography in 251 ophthalmologically normal subjects. Multiple regression analysis for the association of RNFLT with sex, age, axial length, and disc area was performed.

Results  Retinal nerve fiber layer thickness decreased linearly from 125 to 89 μm as the measurement diameter increased (P < .001, mixed linear model). Retinal nerve fiber layer thickness correlated with age in all diameters (partial correlation coefficient [PCC] = −0.40 to −0.32; P < .001) and negatively correlated with disc area in the 2 innermost circles but positively correlated in the 3 outermost circles (PCC = −0.30 to −0.22 and 0.17 to 0.20; P ≤ .005). Sex and axial length did not correlate with RNFLT (P > .08). The decay slope was smallest in the temporal and largest in the nasal and inferior quadrants (P < .001); positively correlated with disc area (PCC = 0.13 to 0.51; P ≤ .04); and negatively correlated with RNFLT (PCC = −0.51 to −0.15; P ≤ .01).

Conclusions  In normal Japanese eyes, RNFLT significantly correlated with age and disc area, but not with sex or axial length. Retinal nerve fiber layer thickness decreased linearly as the measurement diameter increased. The decay slope of RNFLT was steepest in the nasal and inferior quadrants and steeper in eyes with increased RNFLT or smaller optic discs.

Since morphologic changes in the optic disc and the retinal nerve fiber layer (RNFL) often precede observable visual field loss in glaucoma,^1,2 early diagnosis of glaucoma requires an intimate knowledge of the configuration of the optic disc and the distribution of the RNFL thickness (RNFLT) in normal subjects. Though many previous studies have analyzed RNFLT in normal subjects using various imaging techniques, the average values of the 360° peripapillary RNFLT were markedly varied, such as approximately 0.3 mm by confocal scanning laser ophthalmoscopy (Heidelberg Retina Tomograph [HRT]; Heidelberg Engineering, Heidelberg, Germany),^3,4 47 to 95 μm by scanning laser polarimetry (GDx Nerve Fiber Analyzer; Laser Diagnostic Technologies, Inc, San Diego, California),^3-17 and 86 to 128 μm by time-domain optical coherence tomography (TD-OCT).^{3-5,10,12-16,18-43} In these reports, RNFLT was measured at the optic disc edge in HRT, at a diameter of 3.2 mm or a diameter of 1.5 or 1.75 times that of the optic disc in GDx, and at a diameter of 3.40 or 3.46 mm in TD-OCT. Thus, knowledge of the distribution of RNFLT in different diameters around the optic disc should play an important role. However, to our knowledge, the distribution of RNFLT in accordance with the distance and direction from the optic disc has never been investigated in a large normal population.

Spectral-domain optical coherence tomography (SD-OCT) is a newly developed technique.^44-46 Compared with TD-OCT, SD-OCT provides approximately twice the axial resolution and 43 to 100 times the scanning speed and can reveal the 3-dimensional configuration of the retina. To our knowledge, there are no reports of studies of the peripapillary RNFL distribution by SD-OCT in a large group of normal subjects. Further, because a recent study using TD-OCT reported that RNFLT varied significantly among different races or ethnic groups,³⁴ normal RNFLT values should be determined for each ethnic group. The purpose of the present study was to determine the distribution of peripapillary RNFLT in normal Japanese subjects using SD-OCT, as well as the relationship between RNFLT and several demographic factors, such as age, sex, axial length, and disc area.

The data were acquired in 7 institutes in Japan: the University of Tokyo (Tokyo), Kyoto University (Kyoto), Osaka University (Osaka), Niigata University (Niigata), Kanazawa University (Kanazawa), Gunma University (Gunma), and Tajimi Municipal Hospital (Gifu). The study protocol was approved by the institutional review board of each institution and adhered to the tenets of the Declaration of Helsinki. Written informed consent was obtained from each subject after explanation of the study protocol.

Self-reported ophthalmologically healthy subjects at least 20 years of age participated in the study. Ocular examination at the first visit comprised autorefractor keratometry without cycloplegic agents, best-corrected visual acuity measurements, axial length measurement using the IOLMaster (Carl Zeiss Meditec, Dublin, California), slitlamp examination, intraocular pressure measurement using a Goldmann applanation tonometer, dilated funduscopy, and visual field testing using the Humphrey 24-2 Swedish interactive testing algorithm standard strategy (Humphrey Field Analyzer; Carl Zeiss Meditec). Exclusion criteria included contraindication to dilation; intraocular pressure of 22 mm Hg or higher; unreliable Humphrey Field Analyzer results (fixation loss or false-positive or false-negative results >33%); abnormal findings in Humphrey Field Analyzer results suggesting glaucoma according to the Anderson and Patella criteria⁴⁷; history of intraocular surgery; best-corrected visual acuity worse than 20/25; evidence of vitreoretinal diseases; and optic nerve or RNFL abnormality. Eyes with apparent tilted discs that were thought to be congenital deformation⁴⁸ were cautiously excluded, while eyes with slightly tilted discs commonly found in low to moderate myopia were included. After eligibility for study entry was confirmed, all subjects underwent imaging with the 3-dimensional OCT-1000 version 2.13 (Topcon Inc, Tokyo, Japan) after mydriasis. This system uses a superluminescent diode with a center wavelength of 840 nm and a bandwidth of 50 nm as the light source and acquires 27 000 axial scans per second. The OCT data sets were obtained using the concentric scan protocol in which 7 concentric circles with diameters of 2.2, 2.5, 2.8, 3.1, 3.4, 3.7, and 4.0 mm are centered on the gravity center of the optic disc. Using the concentric scan protocol, the image acquisition time was approximately 0.26 second. Each measurement circle included 1024 axial scans. The scan points are closer to each other in the inner circles than in the outer circles, resulting in a higher spatial resolution in the inner circles. For analysis, mean 360° RNFLT was defined as the average of all 1024 points, and the mean quadrant RNFLT was defined as the average of the 256 points of the quadrant. To obtain more accurate circle sizes, the magnification effect in each eye was corrected according to the formula provided by the manufacturer (modified Littman method), based on the refractive error, corneal radius, and axial length, which is basically the same as the original Littman method,⁴⁹ except for small differences in the coefficients of each parameter. To determine the optic disc edge (the inner edge of the scleral rim), 7 points were manually plotted on the optic disc edge of a color fundus photograph obtained by the fundus camera function of the OCT equipment by experienced examiners in each examination site. All of the contour lines of the optic disc were further confirmed by an investigator (A.T.). The center of the optic disc was determined as the barycenter of the closed spline curve fitted to the 7 points, and disc area was determined as the area of the closed curve. During measurements using the concentric scan protocol, the fundus image was continuously monitored to automatically place the concentric circles centered on the thus-determined center of the optic nerve. When eye movement during the measurement was found, the scan was excluded and the measurement was repeated. Criteria for acceptable SD-OCT fundus images were defined as no apparent eye movements during the measurements with a quality factor of more than 60%. The segmentation of the RNFL was performed automatically and an experienced investigator (A.T.) confirmed the segmentations in all scans.

To evaluate the short-term reproducibility of the SD-OCT, measurements were serially repeated 3 times with intervals of less than 1 minute. In some of the subjects, though the numbers were limited because of the time limitation or circumstances of participants, the same protocol was repeated once more by another examiner to test the interoperator reproducibility. Some of the subjects were invited to undergo another set of SD-OCT measurements using the same protocol 1 to 3 months later to test the intervisit reproducibility.

Statistical analysis was performed using SPSS 15.0J (SPSS Institute Inc, Chicago, Illinois). Analysis of variance and the Tukey test were used for the comparison of mean values. Pearson correlation analysis and multiple regression analysis were used to evaluate correlations. Reproducibility was assessed with coefficient of variation. A P value less than .05 was considered statistically significant.

Two hundred fifty-nine subjects were enrolled in the current study, but reliable and acceptable OCT images were not obtained in 9 eyes of 8 subjects. As a consequence, 251 subjects (126 men, 125 women) were included in the analysis (Table 1). The randomly chosen eye of each subject was included in the analyses. Short-term, interoperator, and intervisit reproducibilities were evaluated in 191, 26, and 41 subjects, respectively, and the results are summarized in Table 2. Some of the subjects were overlapped for the 2 or 3 analyses. All coefficients of variation, except those of the innermost circle, were between 1% and 3%, indicating good reproducibility.

Mean 360° RNFLT decreased as the diameter of the measurement circles increased (P < .001) (Table 3). The decay slope was significantly smaller in the temporal quadrant than in the other quadrants (P < .001, analysis of variance and Tukey test), and the slope in the superior quadrant was smaller than those in the inferior and nasal quadrants (P < .001). Age, sex, and axial length did not correlate with the slope (P > .20), whereas disc area significantly correlated with the slope (Table 4), indicating that eyes with a larger optic disc were associated with milder RNFLT decay slopes. Mean total RNFLT also negatively correlated with the slope, indicating that the eyes with greater RNFLT had steeper RNFLT decay slopes.

Mean 360° RNFLT was not different between men and women in any measurement circle after adjusting for age, axial length, and disc area (P > .47). Age was correlated with mean 360° RNFLT in all measurement circles in a simple correlation analysis and multiple regression analysis (P < .001) (Table 5). The rates of RNFL thinning with age were −0.30, −0.27, −0.27, −0.22, −0.19, −0.19, and −0.18 μm/y in the 2.2-, 2.5-, 2.8-, 3.1-, 3.4-, 3.7-, and 4.0-mm measurement circles, suggesting that RNFL thinning was greater in the inner circles. Axial length positively correlated with mean RNFLT in all measurement circles in a simple correlation analysis (P ≤ .01) but not in the multiple regression analysis (P > .08). Optic disc area negatively correlated with mean RNFLT in the 2 innermost circles and positively correlated with mean RNFLT in the outer circles in a simple correlation analysis and multiple regression analysis (P ≤ .04).

Retinal nerve fiber layer thickness measured by the first-generation OCT (OCT1; Zeiss-Humphrey System, Dublin, California and OCT-2000; Humphrey Instruments, San Leandro, California) in normal subjects averaged between 86 and 153 μm.^{5,12,18-31,35} Only 1 study used the second-generation OCT (OCT II; Carl Zeiss Medeitec) and reported a mean value of 105 μm.³² That measured using the third-generation OCT (Stratus OCT; Carl Zeiss Meditec) ranged from 94 to 123 μm.^{3,4,13-16,18,33,34,36-44} If only studies including more than 100 normal subjects are considered, mean RNFLT ranges between 97 and 123 μm.^4,34,41,43 In the present study, using SD-OCT, the mean 360° RNFLT in the 3.4-mm diameter circle was 102 μm in 251 normal subjects. Normal subjects in the present study were Japanese and our results are most similar to the average value (100 μm) obtained using Stratus OCT in 162 Japanese subjects.⁴³

Moreover, in the present study, the distributions of RNFLT around the optic disc were documented at diameters between 2.2 and 4.0 mm. Our results indicated a gradual thinning of the RNFL as the distance from the disc center increased in both the 360° circle and in each of the 4 quadrants (Table 3). Varma et al⁵⁰ reported similar RNFL thinning of the human retina based on histologic analysis, but variations in the decay slope among the different peripapillary directions were not investigated. Gabriele et al⁴⁴ investigated 12 normal eyes using ultra-high resolution SD-OCT and reported that the RNFLT decay slope was significantly smaller in the temporal quadrant than in the other 3 quadrants. In the present study, the RNFLT decay slope in the temporal quadrant was also smaller than that in the other 3 quadrants; furthermore, the RNFLT decay slope in the superior quadrant was smaller than that in the nasal and inferior quadrants. There was no correlation between the slope of RNFLT and age, though correlation between RNFLT and age was stronger in the inner circles. The range of partial correlation coefficients among the circles was not thought large enough to make the correlation between the slope of RNFLT and age significant. Moreover, the current study first documented that the decay slope positively correlated with disc area and negatively correlated with RNFLT, although the causal factors are not clear.

Most studies^{2,6,11,19,25,26,31,33,34,41,51-61} reported a negative correlation between RNFLT and age, indicating that older subjects have thinner RNFL, although some studies^{7,36,50,62-64} failed to show a significant correlation, probably because of the limited numbers of subjects^50,62,64 or the limited range of the subjects' ages.³⁶ Histologic studies^2,52-54,57 revealed significant declines with age in the number of nerve fibers in the optic nerve, with a decay rate of 4000 to 7205 fibers/y. The decay rates in previous TD-OCT studies ranged from 0.16 to 0.44 μm/y.^21,25,34,41 In the present study, significant thinning of the RNFL was associated with an increase in age in the 360° average and all 4 quadrants of all 7 measurement diameters. The rate of the decline in the 3.4-mm diameter circle was 0.19 μm/y, which is consistent with previous reports using Stratus OCT.^34,41 Moreover, in the present study, the correlation between age and RNFLT remained significant after adjusting for sex, axial length, and disc area, indicating that the correlation between age and RNFLT might be attributable not only to the cohort effect in the current population but also to the aging process.

Because axial length and refractive error were closely correlated and a cycloplegic agent was not used, only axial length was adopted in the current analyses. Regarding the correlation between RNFLT and axial length (or refractive error), controversial results have been reported. Using Stratus OCT, several studies showed significant correlation,^34,36,43 while other studies using GDx, OCT-2000, and/or OCT1 failed to find significant correlation.^30,63 In the present study, a significant correlation between RNFLT and axial length was not found after adjusting for age, sex, and disc area. Because axial length differs among different age groups^65,66 and RNFLT is significantly influenced by age, correlations between axial length and RNFLT should be adjusted for age. In some of the previous studies using Stratus OCT,^34,43 the correlation between axial length and RNFLT was still significant after the adjustment for age and other factors. Stratus OCT is not equipped with a magnification correction procedure. In RNFLT measurements without the magnification correction, circle size can be affected by axial length owing to the magnification effect, and circle size is correlated with the RNFLT owing to the slope of the thickness. As a consequence, even if there is no true correlation between axial length and RNFLT, false statistically significant correlation between them could be observed. Though the magnification effect may not be completely corrected, the correction should give better understanding on the true correlation between RNFLT and axial length. Moreover, Leung et al⁶⁷ demonstrated that the correlation between refractive error and disc area determined by Stratus OCT was opposite that determined by the HRT II; however, when the magnification correction was applied to the results of Stratus OCT, the correlation determined by Stratus OCT was similar to that by the HRT II. Thus, the lack of magnification correction in Stratus OCT may at least partly explain the difference in the correlation of RNFLT with axial length between the previous report^34,43 and the present one.

The relation between disc area and the amount of RNFL is controversial. A few histologic studies^50,55,57 with limited numbers of subjects (samples sizes of 10-16) failed to detect a significant correlation. On the other hand, histologic studies with 25 or more subjects (samples sizes of 25 to 56)^53,68,69 demonstrated that larger disc area is significantly related to greater nerve fiber counts. Bowd et al²² reported that there is no correlation between disc area and RNFLT measured by OCT-2000, whereas several studies^34,38,43,70 using Stratus OCT reported a significant correlation (P < .01). In the present study, after adjusting for age, sex, and axial length, there was a positive significant correlation between disc area and RNFLT measured in the 3.4- to 4.0-mm-diameter circles (P ≤ .008), consistent with the previous results obtained with Stratus OCT. On the other hand, opposite correlations were found in the inner circles within a 2.2- to 2.5-mm diameter. As shown by Gabriele et al,⁴⁴ RNFLT initially increased in the area just adjacent to the optic disc edge (usually <2.2 mm diameter) and gradually decreased in the outer diameters. Therefore, the initial increase in RNFLT could affect the results in the 2.2- and 2.5-mm-diameter circles, especially in eyes with larger optic discs, resulting in an inverse correlation between RNFLT and disc area.

Previous studies reported the short-term reproducibility (shown by the coefficient of variation) of RNFLT measurement as 3.85% to 8% as measured using OCT-2000,^21,31,35,71 while 2.2% to 2.8% by Stratus OCT,^42,72 and approximately 2.0% in the peripapillary 3.0- to 6.0-mm area by SD-OCT,⁷³ which is similar to that in the present study. Only 1 study reported the intervisit reproducibility shown by the coefficient of variation using OCT II by 2 operators as 5.3% and 4.3%; however, they did not report interoperator reproducibility.⁷⁴ In the present study, coefficients of variation for short-term, interoperator, and intervisit reproducibilities at a diameter of 3.4 mm were 2.1%, 2.2%, and 2.3%, respectively, suggesting that short-term reproducibility of SD-OCT was as good as that of TD-OCT, and intervisit reproducibilities of the SD-OCT instrument currently used may be better than that of TD-OCT.

The currently used SD-OCT instrument has 2 scan protocols to evaluate the optic disc and peripapillary retina: the concentric scan protocol with a diameter of 2.2 to 4.0 mm and the raster scan protocol, which provides 3-dimensional reconstruction of the retinal structure. In the present study, the results of the concentric protocol were adopted because changes in RNFLT with respect to the distance from the optic disc center could be more easily analyzed and the results more directly compared with the previous literature using Stratus OCT, GDx, and HRT, in all of which peripapillary RNFLT is expressed in a circular fashion.

A limitation of this study is that only Japanese, ophthalmologically normal subjects were included, because Budenz et al³⁴ reported that RNFLT by Stratus OCT was significantly greater in Hispanic and Asian subjects than in white and African American subjects. Another limitation is that measurements of RNFLT with different types of OCT instruments may vary. The recent study by Savini et al⁷⁵ found that Cirrus high-definition OCT (Carl Zeiss Meditec) and RTVue (Optovue Inc, Fremont, California) provide slight but statistically different RNFLT in healthy eyes. Therefore, the results of the present study should better be confirmed in other countries and races and using other SD-OCT instruments.

In conclusion, we report the peripapillary distribution of RNFLT based on SD-OCT in a large number of ophthalmologically normal, Japanese subjects. Retinal nerve fiber layer thickness decreased linearly as the distance from the disc center was increased, with different decay slopes in different directions. Retinal nerve fiber layer thickness was significantly correlated with age and disc area but not with axial length after adjusting for other confounding factors.

Correspondence: Atsuo Tomidokoro, MD, Department of Ophthalmology, University of Tokyo, Graduate School of Medicine, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8655, Japan (tomidokoro-tky@umin.ac.jp).

Submitted for Publication: September 28, 2009; final revision received February 5, 2010; accepted March 6, 2010.

Financial Disclosure: All authors have been paid by Topcon Inc to attend meetings regarding acquisition of the data and Drs Araie and Yoshimura are paid members of the advisory board for Topcon Inc, but none of the authors has a proprietary interest in any products described in the article.

Funding/Support: This work was supported by Grant-in-Aid for Scientific Research by the Ministry of Health, Labor, and Welfare of Japan (H18-Sensory-General-001) and Topcon Inc.