To describe the relationship between peripapillary choroidal thickness and retinal nerve fiber layer (RNFL) thickness in a population-based sample of nonglaucomatous eyes.
Population-based, cross-sectional study.
A total of 478 nonglaucomatous subjects aged over 40 years were recruited from the Singapore Malay Eye Study (SiMES-2). All participants underwent a detailed ophthalmic examination, including Cirrus and Spectralis optical coherence tomography (OCT) for the measurements of RNFL thickness and peripapillary choroidal thickness, respectively. Associations between peripapillary choroidal thickness and RNFL thickness were assessed using linear regression models with generalized estimating equations.
Of the 424 included subjects (843 nonglaucomatous eyes), 60.9% were women, and the mean (SD) age was 66.74 (10.44) years. The mean peripapillary choroidal thickness was 135.59 ± 56.74 μm and the mean RNFL thickness was 92.92 ± 11.41 μm. In terms of distribution profile, peripapillary choroid was thickest (150.04 ± 59.72 μm) at the superior and thinnest (110.71 ± 51.61 μm) at the inferior quadrant, whereas RNFL was thickest (118.60 ± 19.83 μm) at the inferior and thinnest (67.36 ± 11.36 μm) at the temporal quadrant. We found that thinner peripapillary choroidal thickness (PPCT) was independently associated with thinner RNFL thickness globally (regression coefficient [β] = −1.334 μm for per-SD decrease in PPCT, P = .003), and in the inferior (β = −2.565, P = .001) and superior (β = −2.340, P = .001) quadrants even after adjusting for potential confounders.
Thinner peripapillary choroid was independently associated with thinner RNFL globally and in the inferior and superior regions. This structure-structure relationship may need further exploration in glaucomatous eyes prior to its application in clinical settings.
Retinal nerve fiber layer (RNFL) thickness changes are the earliest signs of glaucoma. These precede even optic nerve head (ONH) and visual field changes, making the evaluation of RNFL thickness a crucial assessment in the early diagnosis of glaucoma. Among the various factors associated with the development and progression of glaucoma, vascular and hemodynamic factors have been suggested to play an important role. Studies have now demonstrated vascular insufficiency of the ONH to be an important parameter in the pathogenesis of glaucomatous optic neuropathy. Since RNFL is formed by the expansion of the fibers of the optic nerve, any insufficient blood supply to the ONH could lead to thinner RNFL causing glaucomatous optic neuropathy.
Because of the common source of blood supply to the ONH and peripapillary choroid via the short posterior ciliary arteries, it is likely that a relationship exists between peripapillary choroid and RNFL thickness. However, to date, no studies have explored the quantitative relationship between these parameters in normal subjects, particularly in the general population. Evaluation of the association between peripapillary choroidal thickness and RNFL thickness may help better elucidate the relationship between the structural parameters that may be useful clinically for assessment of ONH damage in glaucoma.
With the recent advancement in imaging technology using spectral-domain optical coherence tomography (SD OCT), in particular the enhanced depth imaging (EDI) technique of SD OCT, objective and quantitative assessment of the peripapillary choroidal thickness is now possible. The purpose of this population-based, cross-sectional study was to evaluate the relationship between peripapillary choroidal thickness and RNFL thickness as measured by SD OCT in a large population sample of nonglaucomatous subjects. We further report the distribution profile of peripapillary choroidal thickness obtained using our automated choroidal segmentation software and RNFL thickness in our population.
Study Population and Design
Subjects of this study were enrolled from the Singapore Malay Eye Study (SiMES), a population-based cohort study of eye diseases in a Malay population aged 40–80 years in Singapore. The baseline examination was conducted between 2004 and 2006 and a follow-up examination of the SiMES participants was conducted between January 2011 and December 2013. For this study, we consecutively recruited 478 subjects from SiMES participants who attended the follow-up examination from February 2012 to July 2013. Written informed consent was obtained from all participants after explanation of the nature and possible consequences of the study. The study adhered to the tenets of the Declaration of Helsinki, and ethics approval was obtained from the Singapore Eye Research Institute Institutional Review Board.
Each study participant underwent a standard ophthalmic examination including measurement of refraction and visual acuity, slit-lamp biomicroscopy, tonometry, pachymetry, perimetry, ocular biometry, fundus examination, and SD OCT imaging. Refraction and corneal curvature were measured using an autokeratorefractometer (Canon RK 5 Auto Ref-Keratometer; Canon Inc Ltd, Tochigiken, Japan). Spherical equivalent (SE) was calculated as the sum of the spherical power and half of the cylinder power. Best-corrected visual acuity (BCVA) was measured monocularly using a logarithm of the minimal angle of resolution (logMAR) chart (Lighthouse International, New York, New York, USA) at a distance of 4 m. Central corneal thickness was measured using an ultrasound pachymeter (Advent; Mentor O & O Inc, Norwell, Massachusetts, USA). Ocular biometry, including axial length (AL), was measured using noncontact partial coherence interferometry (IOL Master V3.01; Carl Zeiss Meditec AG, Jena, Germany). Intraocular pressure (IOP) was measured using Goldmann applanation tonometry (Haag-Streit, Bern, Switzerland) before pupil dilation. Standardized visual field testing was performed with static automated white-on-white threshold perimetry (SITA Fast 24-2, Humphrey Field Analyzer II; Carl Zeiss Meditec, Inc, Oberkochen, Germany). Slit-lamp biomicroscopy (Haag-Streit model BQ-900; Haag-Streit) was performed by the study ophthalmologists to examine the anterior chamber and lens after pupil dilation with tropicamide 1% and phenylephrine hydrochloride 2.5%.
Glaucoma was defined using the International Society of Geographic and Epidemiological Ophthalmology scheme, based on findings from gonioscopy, optic disc characteristics, and visual fields results (as described below).
Visual Field Examination
Standardized visual field testing was performed with static automated perimetry (Swedish Interactive Threshold Algorithm standard 24-2, Humphery Field Analyzer II; Carl Zeiss Meditec, Dublin, California, USA). A visual field was defined as reliable when fixation losses were less than 20%, and false-positive and false-negative rates were less than 33%. A glaucomatous visual field defect was defined as the presence of 3 or more significant ( P < .05) nonedge continuous points with at least 1 at the P < .01 level on the same side of the horizontal meridian in the pattern deviation plot, and classified as “outside normal limits” on the Glaucoma Hemifield Test, confirmed on 2 consecutive visual field examinations.
SiMES is part of the Singapore Epidemiology Eye Diseases (SEED) study. For the purpose of conformity between other studies in SEED, we have used Cirrus HD-OCT for RNFL thickness measurements and Spectralis SD OCT with EDI for choroidal measurements. In addition, we believe that the use of 2 SD OCT machines has its own advantages, as systematic measurement error in 1 machine, if existing, could lead to a biased association between peripapillary choroidal thickness and RNFL thickness, whereas this could be taken care of by the use of 2 machines.
Retinal Nerve Fiber Layer Imaging and Measurement
Cirrus HD-OCT (software version 6.0; Carl Zeiss Meditec, Inc, Dublin, California, USA) was used to measure peripapillary RNFL. After pupil dilation, RNFL scan acquisitions were performed for each participant using an optic disc cube 200 × 200 scan protocol, which generates a cube of data in a 6 mm × 6 mm grid with 200 × 200 axial measurements. In brief, the subject’s pupil was first centered and focused in an iris viewing camera on the acquire screen, and the line scanning ophthalmoscope (LSO) with “auto focus” mode was then used to optimize the view of the retina. The “center” and “enhance” modes were used to optimize the Z -offset and scan polarization, respectively, for the OCT scan to maximize the OCT signal. Rescanning was performed if a motion artifact or saccades through the calculation circle (3.46 mm diameter around the ONH) were detected. The OCT scans were excluded if there was the presence of RNFL or ONH algorithm segmentation failure. All the OCT scans included in the study had signal strength of at least 6, which is considered as acceptable quality. RNFL thicknesses (average, clock hours, and quadrants) were derived automatically from a single scan using the in-built automated software for segmentation and parameter measurements without manual operator adjustment.
Peripapillary Choroidal Thickness Imaging and Measurement
Peripapillary choroid was imaged using the EDI mode of the Spectralis SD OCT. EDI is a method that improves resolution of choroidal details as the zero-delay line with the highest sensitivity is closer to the choroid, and more accurate image acquisition is possible in comparison with those of standard retinal SD OCT methods. Following the Spectralis user manual guidelines, subjects’ keratometry readings and the refraction data were entered into the machine to estimate optical magnification and, therefore, to allow for more accurate comparisons across individuals. The peripapillary region was scanned using a 360 degree, 3.4-mm-diameter circle that was centered on the optic disc, each comprising 100 averaged scans (using the proprietary automatic averaging and eye tracking features of the device). Scans were centered using an internal fixation and centering was confirmed by a scanning laser ophthalmoscope integrated into the instrument.
In our study, the Bruch membrane and choroidal-scleral interface were delineated with the automatic segmentation algorithm developed by Tian and associates. This algorithm demonstrated good consistency with the manual measurements of choroidal thickness (the average of the Dice coefficients over 45 tested images was 90.5% with standard deviation of 3%). The peripapillary choroidal thickness in the optic disc region was automatically measured as the perpendicular distance between the outer portion of the hyperreflective line corresponding to the RPE and the hyporeflective line or margin corresponding to the choroidal-scleral interface at the 12 discrete locations (30 degrees apart) and the 4 quadrants. In our recently published paper using automated choroidal segmentation software we have demonstrated excellent intrasession repeatability (intraclass correlation coefficient ranging from 0.9998 to 0.999) of peripapillary choroidal thickness measurement at all 4 quadrants.
For our analyses, we excluded subjects based on the following criteria: best-corrected logMAR visual acuity >0.30, SE greater than 6 diopter, and clinical features compatible with a diagnosis of glaucoma. The quality of the SD OCT image was assessed prior to the analysis, and images that had motion artifacts or were of insufficient quality (signal strength of <6 for Cirrus OCT and a quality index of <25 decibels for Spectralis OCT, as suggested by the manufacturer, for the image quality assurance) for a reliable determination of RNFL thickness and peripapillary choroidal thickness were excluded. Of the 478 total subjects (947 eyes) examined, 104 eyes were excluded (5 eyes with best-corrected logMAR visual acuity >0.30, 14 eyes with SE greater than -6 diopter, 42 eyes with a diagnosis of glaucoma, and 43 eyes with poor OCT image quality), leaving 843 nonglaucomatous eyes for final analysis.
Mean and standard deviation (SD) of both peripapillary choroidal thickness and RNFL thickness was calculated in all subjects for clock hours and 4 quadrants. Associations of peripapillary choroidal thickness (independent variable of interest) with RNFL thicknesses (dependent variable) were assessed using linear regression. Generalized estimating equations (exchangeable correlation matrix) were used to account for the correlation between pairs of eyes for each individual. Factors such as age, sex, AL, IOP, diabetic retinopathy, and age-related macular degeneration were included in the multivariate model to adjust for potential confounding. Statistical significance was set at P < .05 unless otherwise indicated. The data were analyzed with MedCalc version 12.3 (Medcalc Software, Ostend, Belgium) and SPSS version 20.0 (SPSS, Inc, Chicago, Illinois, USA).
A total of 843 eyes from 424 subjects were included in the study. Of the 424 subjects, 363 (85.6%) did not have any eye diseases; the remaining 61 (14.4%) had eye diseases including diabetic retinopathy (8.5%) and early (5.2%) and late (0.6%) age-related macular degeneration. The included participants’ mean age was 66.74 ± 10.44 years and 258 (60.9%) participants were female. The clinical characteristics of the included and excluded eyes are shown in Table 1 . Compared to the eyes included in the analysis, excluded eyes were more myopic and had poor BCVA.
(n = 843 Eyes)
(n = 61 Eyes)
|P Value a|
|Axial length, mm||23.56 (0.98)||24.27 (1.92)||.198|
|Anterior chamber depth, mm||3.14 (0.38)||3.00 (0.42)||.185|
|Corneal curvature, mm||7.67 (0.25)||7.64 (0.24)||.876|
|Spherical equivalent, D||0.23 (1.57)||−3.19 (5.54)||.002|
|BCVA, logMAR||0.28 (0.45)||0.55 (0.78)||.028|
|Central corneal thickness, μm||539.72 (32.64)||536.10 (32.16)||.063|
|Intraocular pressure, mm Hg||14.30 (3.17)||15.48 (4.92)||.150|
|Average RNFL thickness, μm||92.92 (11.41)||93.11 (11.09)||.923|
a P value was obtained with generalized estimating equation.
Table 2 presents the distribution of mean peripapillary choroidal thickness and RNFL thickness measured at 12 clock hours and 4 quadrants (superior, nasal, inferior, and temporal). The average peripapillary choroidal thickness was 135.59 ± 56.74 μm and the average RFNL thickness was 92.92 ±11.41 μm. There are variations in the topographic profile of peripapillary choroidal thickness and RNFL thickness among clock hours and different quadrants. Peripapillary choroid was thickest (150.04 ± 59.72 μm) at the superior quadrant and thinnest (110.71 ± 51.61) at the inferior quadrant, whereas RNFL was thickest (118.60 ± 19.83 μm) at the inferior and thinnest (67.36 ± 11.36) at the temporal quadrant.
|Measurement Location||Peripapillary Choroidal Thickness (μm)||Retinal Nerve Fiber Layer Thickness (μm)|
|Mean (SD)||Mean (SD)|
|1||138.61 (77.01)||56.95 (11.08)|
|2||147.09 (70.88)||81.10 (16.00)|
|3||148.38 (65.03)||115.72 (23.97)|
|4||150.46 (62.21)||116.95 (26.79)|
|5||151.06 (59.39)||115.26 (24.91)|
|6||151.68 (61.06)||81.07 (16.87)|
|7||145.17 (61.97)||57.38 (11.06)|
|8||133.58 (58.49)||67.01 (14.72)|
|9||115.32 (53.67)||113.61 (30.65)|
|10||102.01 (51.07)||127.30 (28.03)|
|11||113.25 (57.11)||115.00 (30.08)|
|12||129.24 (68.31)||67.53 (14.04)|
|Superior||150.04 (59.72)||116.05 (18.42)|
|Nasal||143.12 (58.46)||69.69 (11.28)|
|Inferior||110.71 (51.61)||118.60 (19.83)|
|Temporal||138.47 (68.65)||67.36 (11.36)|
|Average thickness||135.59 (56.74)||92.92 (11.41)|
Table 3 shows the linear regression analyses of the associations of peripapillary choroidal thickness (exposure variable of interest) by locations evaluated against RNFL thicknesses (dependent variable) from the same location ( Figure ) to calculate regression coefficients (β). In Model 1, including age, sex, and AL, we found that thinner peripapillary choroidal thickness was independently associated with thinner RNFL thickness globally (β = −1.364 μm for per-SD decrease in peripapillary choroidal thickness, P = .002) and in the inferior (β = −2.539, P = .001) and superior (β = −2.492, P = .001) quadrants. In Model 2, including age, sex, AL, IOP, diabetic retinopathy, and age-related macular degeneration, thinner peripapillary choroidal thickness was independently associated with thinner RNFL thickness globally (β = −1.329 μm for per-SD decrease in peripapillary choroidal thickness, P = .003) and in the inferior (β = −2.566, P < .001) and superior (β = −2.348, P = .001) quadrants. In order to control for potential confounding effect from OCT signal strengths, we further included signal strength from both Cirrus and Spectralis OCT in our regression analyses. The results remained similar after further adjustment for signal strength (Model 3 in Table 3 ). However, in all models, the association was not present in the temporal quadrant.