Correct Calculation Circle Location of Optical Coherence Tomography in Measuring Retinal Nerve Fiber Layer Thickness in Eyes with Myopic Tilted Discs

Abstract

Purpose.: Retinal nerve fiber layer (RNFL) thickness measured by optical coherence tomography (OCT) can be unreliable in the presence of a myopic tilted disc. The purpose of this study was to investigate the influence of the calculation circle location of spectral domain (SD)-OCT when measuring RNFL thickness in healthy subjects with myopic tilted disc.

Methods.: Sixty-nine eyes with myopic tilted disc and no other ocular abnormalities were analyzed. In all eyes, RNFL thickness was measured twice: first, along a calculation circle determined by a built-in algorithm (referred to as circle 1) and then along a manually positioned calculation circle based on the contours of the neural canal opening (NCO) (referred to as circle 2). The mean number of clock hours below normal at the 5% level and the proportion of eyes with abnormally thin clock hours at the 5% level were compared between the two methods.

Results.: The mean number of clock hours below normal at the 5% level was significantly lower in circle 2 (0.84 ± 1.31) than in circle 1 (1.42 ± 1.43; P < 0.05). The proportion of eyes with abnormally thin clock hours at the 5% level was significantly lower (P < 0.001) in circle 2 (42%) than in circle 1 (69.9%).

Conclusions.: In subjects with myopic tilted disc, RNFL thickness measurements along the calculation circle based on the contours of the NCO seem to be more comparable to the normative database of the SD-OCT than does the automatically determined scan position.

Optical coherence tomography (OCT) is a useful instrument for the diagnosis and follow-up of glaucoma, owing to its excellent ability to quantitatively assess the thickness of the peripapillary retinal nerve fiber layer (RNFL). ^1,2  

The newly developed spectral domain (SD)-OCT offers higher axial resolution and scanning speed (axial resolution, 5 μm; 27,000 axial scans per second) than does time-domain OCT. ³ In addition, SD-OCT automatically identifies a disc center and places a calculation circle around the optic disc. ⁴ Therefore, it can offer more accurate and reproducible results. ⁵  

However, the performance of OCT in evaluating subjects with myopic eyes remains uncertain. ⁶ Some recent studies ^6,7 mention that false-positive errors can arise when analyzing the topographic profile of RNFL thickness in myopic eyes. Furthermore, the presence of a myopic tilted disc can make it difficult to determine the optic disc margin, thus hindering RNFL analysis by OCT. ^{7 –9}  

Proper RNFL assessment of eyes with myopic tilted disc is important because myopia is an independent risk factor for glaucoma and because a tilted disc is more prevalent in highly myopic eyes. ^10,11  

We hypothesized that unreliable RNFL measurements in eyes with myopic tilted disc result from an inappropriate location of the calculation circle from OCT. The purpose of this study was to investigate the influence of SD-OCT calculation circle location on the measurement of RNFL thickness in eyes with myopic tilted disc. 

Materials and Methods

Subjects

Study subjects included healthy young Korean male subjects with no known ocular abnormalities who visited for routine eye examinations or desired a prescription for glasses at Kangdong Sacred Heart Hospital from April 2008 through December 2009. 

The procedures used in this study conformed to the guidelines of the Declaration of Helsinki. This study was approved by the institutional review board of Kangdong Sacred Heart Hospital. Informed consent was obtained from each subject. 

All subjects underwent a full ophthalmic examination, including measurements of visual acuity (BCVA) and refraction by automatic refractometer (model R-F10; Canon Inc., Tokyo, Japan), intraocular pressure by Goldmann applanation tonometry, slit lamp examination, A-scan ultrasound biometry, dilated fundus examination, gonioscopic examination by a Goldmann three-mirror lens, color disc and red-free digital fundus photography (CF-60UD camera; Canon, Tokyo, Japan), and standard automated perimetry (Humphrey Field Analyzer II 750; central 24-2 pattern of the standard Swedish Interactive Threshold Algorithm; Carl Zeiss Meditec, Dublin, CA). 

Individuals exhibiting any of the following conditions were excluded: BCVA < 20/20; a history of ocular trauma or intraocular or refractory surgery; any active infection of the anterior or posterior segments; evidence of retinal disease, including diabetic retinopathy, macular edema, or other vitreoretinal disease; any history of optic nerve disease, including nonglaucomatous optic neuropathy or optic disc anomaly; presence or history of ocular hypertension (IOP > 21 mm Hg); a closed or occludable angle on gonioscopic examination; presence or history of disc hemorrhage in either eye; optic disc with glaucomatous (i.e., asymmetric cup-disc ratio or focal notching or narrowing of neuroretinal rim) or nonglaucomatous changes; RNFL atrophy on the red-free fundus photograph; evidence of a reproducible visual field defect (Glaucoma Hemifield Test results outside normal limits, pattern standard deviation with P < 5% or a cluster of three or more nonedge points on the pattern deviation plot in a single hemifield with P < 5%, one of which must have a P < 1%) in either eye; unreliable visual field test results (>15% false positives or >20% fixation loss); unusable OCT scans because of poor signal strength (<7) or algorithm failure in which the boundaries for the RNFL were not delineated correctly; inadequate color disc and red-free fundus photography quality; or a history of systemic disease, such as diabetes, leukemia, AIDS, uncontrolled systemic hypertension, or multiple sclerosis. 

The eyes that satisfied the study criteria were divided into two groups: (1) the myopic tilted-disc group, which included the eyes with various degrees of myopia and a temporally tilted disc, and (2) the control group, which included the remaining eyes without a tilted disc. 

Tilted Optic Disc Assessment

The pupil was fully dilated with tropicamide (0.5%) and phenylephrine (0.5%). We took 30° color photographs of the optic disc with the digital fundus camera (CF-60UD digital camera; Canon) and reviewed them on LCD monitors. Tilted optic discs were screened by a glaucoma specialist (YCY) who was masked to the subject's clinical data. The specialist used binocular optic disc evaluation and color disc photographs of the subjects with myopia to perform evaluations. The criteria for a tilted disc ¹² included the following: the presence of a temporal scleral crescent (or peripapillary atrophy [PPA]); the appearance of tilting in the disc surface in three-dimensional viewing; a steeper nasal neuroretinal rim surface and a neuroretinal rim surface on the side of the temporal PPA that was less steep than the retinal surface; and an overall more oval disc shape. 

To confirm the tilted discs ascertained based on the stereoscopic evaluation, another glaucoma specialist (JKC) re-examined the tilted discs by measuring the tilt ratio, which was calculated by dividing the minimum disc diameter by the maximum disc diameter. A tilted optic disc was defined as an optic disc with a tilt ratio ≤0.80, as in previous studies. ¹¹ With the disc photograph projected on an LCD monitor, minimum and maximum disc diameters were measured in pixels, and the tilt ratio was calculated (ruler tool in Photoshop CS3; Adobe Systems, San Jose, CA). The measurements were made from the disc margins defined as the inner border of the peripapillary scleral ring. 

Optical Coherence Tomography

The eyes of the subjects who satisfied the study criteria after pharmacologic dilation of the pupil were scanned with SD-OCT (Cirrus HD-OCT; software version 4.5.1.11; Carl Zeiss Meditec, Dublin, CA) by a single expert examiner. After the subject was seated and aligned properly, three 200 × 200-cube optic disc scans were obtained per eye. A built-in algorithm in the HD-OCT located the center of the optic disc, even if it was not accurately centered in the scan image. The HD-OCT identified the disc center by finding a dark spot near the center of the scan that had a shape and size consistent with a range of optic disc shapes and sizes. ⁴ A calculation circle 3.46 mm in diameter consisting of 256 (scans 0–255) A-scans was then automatically positioned around the optic disc. If the signal strength of the scan was less than 7, the scan was discarded and the scanning procedure was repeated. If the calculation circle centered on the optic disc passed the area of the PPA, the eye was excluded from further analysis. The RNFL thicknesses at the 256 points of the RNFL thickness profile, mean RNFL thicknesses in the entire circle and each quadrant, and clock hours were recorded. Clock-hour RNFL thickness was recorded on the basis of right-eye orientation. The superior clock hour was 12 o'clock, and the others were assigned accordingly, clockwise and counterclockwise in the right and left eyes, respectively. 

In this study, RNFL thickness measurements were performed twice in the tilted-disc group. After automated RNFL analysis by the HD-OCT built-in algorithm (Fig. 1A), we manually placed the calculation circle on a new location centered on the margin involving the temporal edge of the PPA and nasal scleral lip on a magnified HD-OCT fundus image (Fig. 1B). The Cirrus HD-OCT allows the controlled shifting of the 3.46-mm calculation circle with respect to the center of the optic nerve head and does not require the user to obtain multiple scans. ¹³ This procedure was performed by the authors in a blinded fashion. In the present study, RNFL thicknesses measured by the automatically placed and manually positioned calculation circles are termed circle 1 and circle 2, respectively. 

Figure 1.

 

An example of circle 1 (A) and circle 2 (B). (A) A Cirrus HD-OCT fundus image with the 3.46-mm scan circle (white circle) that was automatically positioned around the optic disc (dotted black circle). (B) A manually positioned calculation circle location (black circle) centered on the margin involving the temporal edge of the peripapillary atrophy and nasal scleral lip (dotted black circle) that was displaced from that of circle 1 (white circle).

Data Analysis

The clinical characteristics of the tilted disc and control groups were analyzed using the paired t-test (all data analyses, SPSS, ver. 12.0; SPSS, Chicago, IL). The paired t-test was also used to compare the RNFL thicknesses of circles 1 and 2 in the tilted-disc group. The McNemar test was used to compare the probability of abnormal OCT parameters at the 5% level, the mean number of clock-hour sectors below the 5% level, and the proportion of eyes with abnormally thin clock-hour sectors at the 5% level with respect to the built-in RNFL normative database between circles 1 and 2. Statistical significance was set at P < 0.05. The intraclass correlation coefficient (ICC) was computed to determine the reproducibility of the manual placement of the calculation circle for measuring peripapillary RNFL thickness between the two examiners (the authors). The ICC represents concordance where 1 is perfect agreement and 0 is no agreement at all. The coefficient of variation (CV), calculated as the square root of the variance divided by the mean peripapillary RNFL thickness of two measurements, and test–retest variability (TRV), calculated as twice the square root of the variance between two measurements, were also computed to determine the agreement between circle 2 measurements. In addition, linear-weighted kappa coefficients (κ) were calculated as indicators of agreement between the two examiners on the color-code classification, which was based on the age-matched RNFL normative data from Cirrus HD-OCT (red < 1%; 1% ≤ yellow < 5%; 5% ≤ green ≤ 95%; white > 95%). 

Results

This study initially included 504 eyes from 252 healthy young Korean male subjects who had no abnormal ophthalmic findings. Twelve eyes were excluded from further analysis because of inadequate of stereo disc photography quality (n = 4) or unacceptable HD-OCT scans (n = 8), leaving a final sample of 492 eyes of 252 subjects. Subjects were included by consensus between the two glaucoma specialists (the authors). Among the 492 eyes, 69 satisfied the criteria for a tilted disc on stereoscopic examination. Because the tilt ratios of all 69 eyes were below 0.8 (<0.70: n = 15; 0.70–0.74: n = 17; and 0.75–0.80: n = 37) and the mean tilt ratio was 0.73 ± 0.05, we enrolled the 69 eyes in the tilted-disc group and the remaining 423 without tilted discs in the control group. 

Subject characteristics are summarized in Table 1. The tilted-disc group exhibited significantly longer axial lengths (P < 0.001) and a higher degree of myopia (spherical equivalent; P < 0.001) than the control group. There was a significant correlation between axial length and spherical equivalent (r = −0.639, P < 0.001). In addition, the tilted-disc group exhibited thinner whole-scan RNFL thicknesses (P < 0.001) than did the control group. Regarding quadrant analysis, the tilted-disc group had thinner RNFL thicknesses in the superior (P < 0.001), inferior (P < 0.001), and nasal (P < 0.001) quadrants; however, the RNFL was thicker in the temporal quadrant (P < 0.001) than that in the control group. There were no significant differences in age (P = 0.649) or intraocular pressure (P = 0.788). 

Table 1.

 

Clinical Characteristics of Control and Tilted-Disc Groups

	Control Group (423 Eyes)	Tilted-Disc Group (69 Eyes)	P *
Age, y	21.4 ± 1.7 (19 to 28)	21.5 ± 1.5 (19 to 25)	0.649
IOP, mm Hg	15.5 ± 2.7 (7 to 21)	15.6 ± 3.0 (9 to 21)	0.788
Spherical equivalent, D	−2.2 ± 2.1 (−11.0 to +4.1)	−4.7 ± 2.5 (−10.0 to −0.1)	<0.001
Axial length, mm	24.5 ± 1.2 (21.4 to 28.8)	25.8 ± 1.2 (23.1 to 28.1)	<0.001
Average RNFL thickness, μm
Whole scan	99.0 ± 9.1	94.4 ± 9.8	<0.001
Superior quadrant	128.9 ± 15.8	118.8 ± 14.7	<0.001
Inferior quadrant	126.8 ± 16.6	116.5 ± 16.9	<0.001
Temporal quadrant	71.2 ± 11.1	80.4 ± 16.5	<0.001
Nasal quadrant	68.9 ± 11.0	61.8 ± 13.3	<0.001

 Data are expressed as the mean ± SD (range).

*  Paired t-test.

The reproducibility of the manual placement of calculation circles centered on the margin involving the temporal edge of the PPA and nasal scleral lip (circle 2) is summarized in Table 2. The strength of agreement for the whole-scan thickness measurement and for the quadrant and clock-hour RFNL measurements between the two examiners was excellent (>0.80). ¹⁴ Furthermore, the reproducibility of all color-coded classifications was substantial to almost perfect (from 0.702–1.000). ¹⁵ Because the determination of the location of circle 2 was highly reproducible, we intentionally selected one of the two authors' independent measurements of circle 2 for analysis in this study. 

Table 2.

 

Reproducibility of Peripapillary RNFL Thickness Measurement (ICC, TRV, and CV) and Its Color-Coded Classification (κ) in Circle 2, Manually Positioned by Two Examiners

Parameter	ICC	TRV (μm)	CV (%)	κ*
Whole scan	0.9961 (0.9938–0.9976)	1.2	0.6	0.908 (0.780–1.000)
Superior quadrant	0.9508 (0.9219–0.9692)	6.8	2.8	0.894 (0.744–1.000)
Inferior quadrant	0.9777 (0.9642–0.9861)	4.7	2.0	0.926 (0.835–1.000)
Temporal quadrant	0.9659 (0.9455–0.9788)	4.2	3.3	0.809 (0.580–1.000)
Nasal quadrant	0.9486 (0.9185–0.9678)	5.2	3.7	0.820 (0.581–1.000)
12 o'clock	0.9529 (0.9252–0.9705)	10.9	4.4	1.000 (1.000–1.000)
1 o'clock	0.9531 (0.9253–0.9707)	10.4	4.7	0.752 (0.567–0.938)
2 o'clock	0.9717 (0.9548–0.9824)	5.9	3.9	0.917 (0.801–1.000)
3 o'clock	0.9252 (0.8821–0.9530)	6.3	5.4	1.000 (1.000–1.000)
4 o'clock	0.9649 (0.9439–0.9781)	4.5	3.9	0.777 (0.573–0.981)
5 o'clock	0.9702 (0.9524–0.9815)	5.9	3.2	0.702 (0.535–0.869)
6 o'clock	0.8353 (0.7471–0.8946)	20.5	8.1	0.812 (0.634–0.991)
7 o'clock	0.9773 (0.9636–0.9859)	7.1	2.5	1.000 (1.000–1.000)
8 o'clock	0.9226 (0.8782–0.9513)	7.7	5.3	0.841 (0.627–1.000)
9 o'clock	0.8719 (0.8013–0.9186)	5.7	5.2	0.742 (0.398–1.000)
10 o'clock	0.9711 (0.9538–0.9820)	5.5	3.5	0.919 (0.762–1.000)
11 o'clock	0.9704 (0.9526–0.9816)	7.5	2.9	0.883 (0.735–1.000)

 95% CI in parentheses. ICC, intraclass correlation coefficient; TRV, test-retest variability; CV, coefficient of variation

*  Linear-weighted κ coefficients.

Table 3 compares the thicknesses of the OCT parameters for circles 1 and 2. There were no significant differences in the whole-scan RNFL thickness between circles 1 and 2 (P = 0.109). For quadrant measurements, the RNFL thickness of circle 2 was significantly thicker in the superior (P = 0.032), inferior (P < 0.001), and nasal (P = 0.006) quadrants, but thinner in the temporal (P < 0.001) quadrant than those of circle 1. On the basis of the clock-hour analysis, the RNFL thicknesses in the 7-, 8-, 9-, 10-, and 11-o'clock sectors were significantly thinner (all P < 0.001), whereas the 1-, 2-, 5-, 6-, and 12-o'clock sectors were significantly thicker (all P < 0.001) in circle 2 than those in circle 1. The clock-hour thicknesses in the 3- and 4-o'clock sectors were also thicker in circle 2 than those in circle 1; however, these differences were not statistically significant (P = 0.64 and 0.254, respectively). 

Table 3.

 

Comparison of RNFL Thicknesses between Circles 1 and 2

	Circle 1*	Circle 2†	P ‡
Whole scan	94.4 ± 9.8	94.0 ± 9.5	0.109
Superior quadrant	118.8 ± 14.7	120.8 ± 15.4	0.032
Inferior quadrant	116.5 ± 16.9	120.8 ± 15.8	<0.001
Temporal quadrant	80.4 ± 16.5	69.5 ± 11.0	<0.001
Nasal quadrant	61.8 ± 13.3	64.8 ± 11.5	0.006
12 o'clock	118.4 ± 23.5	123.8 ± 24.9	0.004
1 o'clock	100.0 ± 23.2	110.0 ± 24.5	<0.001
2 o'clock	69.9 ± 14.4	76.8 ± 17.6	<0.001
3 o'clock	58.3 ± 12.0	58.6 ± 11.5	0.640
4 o'clock	56.4 ± 16.4	58.4 ± 12.3	0.254
5 o'clock	81.7 ± 15.6	93.1 ± 17.7	<0.001
6 o'clock	116.2 ± 26.5	128.1 ± 26.5	<0.001
7 o'clock	151.6 ± 27.1	141.3 ± 23.1	<0.001
8 o'clock	88.2 ± 20.9	73.3 ± 13.7	<0.001
9 o'clock	63.3 ± 12.0	55.2 ± 8.0	<0.001
10 o'clock	92.0 ± 22.6	80.2 ± 15.7	<0.001
11 o'clock	138.2 ± 22.6	128.9 ± 12.9	<0.001

 Data are expressed as mean micrometers ± SD.

*  Calculation circle automatically placed by the Cirrus HD-OCT built-in algorithm.

†  Manually positioned calculation circle centered on the margin involving the temporal edge of the peripapillary atrophy and nasal scleral lip, which corresponds to the neural canal opening.

‡  Paired t-test.

Table 4 shows the comparison of the probability of abnormal OCT parameters at the 5% level (i.e., the false-positive rate) between circles 1 and 2, with respect to the built-in RNFL normative database. In circle 2, the false-positive rate was lower in the nasal and inferior quadrants but higher in the temporal quadrant than in circle 1. However, a statistically significant change was observed only in the nasal quadrant (P = 0.0078). Regarding the clock-hour analysis, the false-positive rates in the 1-, 2-, 5-, and 6-o'clock sectors were significantly lower in circle 2 than in circle 1 (P = 0.0078, 0.0039, 0.0001, and 0.002, respectively). 

Table 4.

 

Comparison of the Probability of Abnormal OCT Parameters at the 5% Level (False-Positive Rate) between Circles 1 and 2 with Respect to the Built-in RNFL Normative Database

	Circle 1*	Circle 2†	Circle 1 vs. Circle 2	P ‡
Whole scan	13.0	8.7	4.3	0.25
Superior quadrant	5.8	5.8	0	1.0
Inferior quadrant	20.3	13.0	7.3	0.0625
Temporal quadrant	0.0	1.5	−1.5	1.0
Nasal quadrant	21.7	10.1	11.6	0.0078
12 o'clock	4.3	1.4	2.9	0.5
1 o'clock	21.7	11.6	10.1	0.0078
2 o'clock	27.5	15.9	11.6	0.0039
3 o'clock	4.3	2.9	1.4	1.0
4 o'clock	18.8	14.5	4.3	0.25
5 o'clock	34.8	13.0	21.8	0.0001
6 o'clock	24.6	10.1	14.5	0.002
7 o'clock	1.4	1.4	0.0	NA
8 o'clock	0.0	0.0	0.0	NA
9 o'clock	0.0	1.4	−1.4	1.9
10 o'clock	2.9	5.8	−2.9	0.5
11 o'clock	2.9	7.2	−4.3	0.25

 Data are expressed as percentages. NA, not applicable.

*  Calculation circle automatically placed by the Cirrus HD-OCT built-in algorithm.

†  Manually positioned calculation circle centered on the margin involving the temporal edge of the peripapillary atrophy and nasal scleral lip, which corresponds to the neural canal opening.

‡  McNemar test.

Table 5 shows the mean number of clock-hour sectors below normal at the 5% probability level and compares the proportion of eyes with abnormally thin clock-hour sectors at the 5% probability level with the built-in RNFL normative database for circles 1 and 2. The mean number of abnormal clock-hour sectors at the 5% level was significantly lower in circle 2 (0.84 ± 1.31) than in circle 1 (1.42 ± 1.43; P < 0.05). 

Table 5.

 

Comparison of the Number of Clock-Hour Sectors Below Normal at the 5% Probability Level and the Proportion of Eyes with Abnormally Thin Clock-Hour Sectors at the 5% Probability Level between Circles 1 and 2

	Circle 1*	Circle 2†	P
Number of abnormal clock-hour sectors, mean ± SD	1.42 ± 1.43	0.84 ± 1.31	<0.05‡
Proportion of eyes with abnormal clock-hour sectors, %
One or more clock-hours	48/69 (69.9)	29/69 (42.0)	<0.001§
Two or more clock-hours	27/69 (39.1)	16/69 (23.2)	<0.05§
Three or more clock-hours	13/69 (18.8)	6/69 (8.7)	0.07§

*  Calculation circle automatically placed by the Cirrus HD-OCT built-in algorithm.

†  Manually positioned calculation circle centered on the margin involving the temporal edge of the peripapillary atrophy and nasal scleral lip, which corresponds to the neural canal opening.

‡  Paired t-test.

§  McNemar test.

Furthermore, the proportion of eyes that had abnormal clock-hour sectors at the 5% level was lower in circle 2 than in circle 1. Specifically, the proportion of eyes that had one or more abnormal clock-hour sectors in which the RNFL thickness was below normal at the 5% probability level in circle 2 (29/69, 42.0%) was significantly lower (P < 0.001) than that in circle 1 (48/69, 69.9%). In addition, 27 (39.1%) of 69 eyes exhibited 2 or more abnormally thin clock-hour sectors with circle 1, while only 16 (23.2%) of 69 exhibited such findings with circle 2; the differences in these proportions were statistically significant (P < 0.05). Furthermore, 13 (18.8%) of 69 eyes exhibited three or more abnormally thin clock-hour sectors with circle 1 compared to 6 (8.7%) of 69 with circle 2; however, these differences were not statistically significant (P = 0.07). 

Discussion

In this study, we evaluated the peripapillary RNFL thicknesses of young healthy subjects with myopic tilted disc to determine the possible cause of unreliable RNFL measurements in myopic tilted disc by OCT. We excluded glaucomatous eyes by enrolling only young subjects with a symmetric cup-disc ratio and without focal notching or narrowing of the neuroretinal rim and no reproducible visual field defect or RNFL atrophy on red-free fundus photography. In addition, all subjects were enrolled based on a consensus between two glaucoma specialists (the authors). 

We found that the mean peripapillary RNFL thickness value on the whole scan was smaller in the tilted-disc group that that in the controls. This finding can be explained by previous studies that reported an association between thinner RNFL values with increasing axial length and higher degree of myopia, ^16,17 because the tilted-disc group in our study had a significantly higher degree of myopia and longer axial length than the control group (Table 1). Furthermore, the RNFLs in eyes with tilted discs were significantly thinner in the nontemporal quadrants and thicker in the temporal quadrant compared to that in the controls. Likewise, previous studies found that highly myopic eyes may have a RNFL topographic profile that differs from that of nonmyopic eyes, ^4,6,18 and this may have implications in the use of OCT for detecting glaucomatous damage because the current OCT technology determines the damage by comparing the RNFL thickness with the normative database on a sector-by-sector basis (quadrant or clock hour). 

Lower average peripapillary RNFL thickness and different RNFL topographic profile with more pronounced thinning in the nontemporal sectors in eyes with myopic tilted disc may lead to false-positive errors, especially in the nontemporal sectors. ⁶ We therefore assumed that unreliable RNFL measurements in eyes with myopic tilted disc result from false-positive errors and may be related to inappropriate location of the calculation circle from OCT. 

In this study, we introduced a novel reference plane for SD-OCT that significantly reduced the mean number of abnormal clock-hour sectors at the 5% level (P < 0.05) as well as the proportion of eyes with abnormally thin clock-hour sectors at the 5% level (P < 0.001) compared with the Cirrus normative database in normal subjects with myopic tilted disc. RNFL values from the calculation circle centered on the margin involving the temporal edge of the PPA and nasal scleral lip are very comparable to the normative database of the SD-OCT. This finding is an intriguing one that has never been reported. 

The theoretical background of our reference plane was based on the following studies. Strouthidis et al. ¹⁹ evaluated the three-dimensional anatomy of the optic nerve head in highly myopic human eyes by using their own technique. On the basis of the finding that the NCO, which is the anatomic opening in Bruch's membrane (also referred to as Bruch's membrane opening), is potentially the first aperture through which the retinal ganglion cell axons travel as they pass through the scleral canal, ^20,21 they suggest that the NCO is a suitable reference plane for SD-OCT imaging and an important landmark for optic nerve head and peripapillary RNFL quantification. In their study of the three-dimensional anatomy of human myopic optic discs, the temporal NCO coincided with the termination of the retinal pigment epithelium at the temporal edge of the PPA; furthermore, the nasal NCO corresponds to the nasal scleral lip. ²²  

Kotera et al. ²² also highlight the circumstances in which the termination of the retinal pigment epithelium identified by SD-OCT coincides with the termination of visible retinal pigment at the edge of the PPA. In addition, Nakazawa et al. ²³ reported that the distance between the fovea centralis and optic disc gradually changes in proportion to the progression of myopia, which is regarded to be representative of optic disc deviation; however, the temporal margins of the crescent always maintain the same distance from the fovea centralis and correspond to the original optic disc. 

On the basis of the results of those studies, we assumed that the temporal edge of the temporal PPA and nasal scleral lip that corresponded to the NCO might be the original optic disc margin and appropriate landmarks for the OCT calculation circle location in myopic subjects with tilted discs. 

When using the Cirrus HD-OCT, any shift in the location of the calculation circle that brings a sector closer to the disc margin causes an increase in thickness in the region and a decrease in the opposite region. ¹³ We thought the calculation circle determined by the Cirrus HD-OCT built-in algorithm (circle 1) deviated nasally because we designated the temporal edge of the temporal PPA as a landmark for the temporal side of the calculation circle instead of the temporal scleral lip (Fig. 1). Nasal displacement of the scan brings the thickened bundles of the RNFL into the temporal region and increases the thickness in that region and decreases the thickness in the nontemporal region. 

In addition, a thicker temporal quadrant and thinner nontemporal quadrant in this study can be explained by temporal deviations in the peak RNFL thickness. In myopic eyes, the location of the maximum RNFL thickness for the superotemporal and inferotemporal RNFL humps (referred to as the superior and inferior peaks, respectively) is closer to the temporal horizon than that in nonmyopic eyes. As the axial length increases, the retina is dragged toward the temporal horizon, and the RNFL layers are compressed against the bundles originating from the opposite hemisphere at the horizontal raphe. This effect results in the thickening of the RNFL in the temporal quadrant. In contrast, the other quadrant becomes thinner as it is stretched. ^6,24 This result could be applicable to our study, because subjects with myopic tilted discs had significantly longer axial lengths and higher degrees of myopia than those in the control group (Table 1). 

We hypothesized that as the calculation circle is moved temporally, the nasal displacement of the scan can be corrected, and the location of the superior and inferior peaks can be widened; this would result in the thickening of the nontemporal quadrants and clock hours ¹³ and could reduce false-positive errors. The peak locations of the superior/inferior RNFL were evaluated by the RNFL TSNIT (temporal, superior, nasal, inferior, temporal) curve on HD-OCT. The peak locations appeared at 1 of the 256 points (ranging from 0 to 255) provided by the Cirrus HD-OCT algorithm. We converted the determined peak location into the angle (in degrees) between an imaginary horizontal line (Fig. 2A) and the superotemporal or inferotemporal hump, by multiplying it by 360/256. For example (serial number 59), the superior peak location of 37 in the TSNIT curve was translated to 52.0° (37 × 360/256), and the inferior peak location of 213 in the ISNIT curve was translated to 60.5° [(256–213) × 360/256]; Figs. 2B, 2C). This result means that the thickest superior RNFL was located 52.0° away from the temporal horizontal meridian, and the inferior peak was located 60.5° away from the temporal horizontal meridian. 

Figure 2.

 

An example (serial number 59) showing that temporal displacement of the calculation circle can widen the peak distance. (A) Fundus photography of the optic disc. Dotted black line: the imaginary horizontal meridian. (B) The RNFL TSNIT curve of the Cirrus HD-OCT. The peak locations at the superior and inferior area were 37 and 213 in circle 1, and 48 and 209 in circle 2, respectively. (C) The peak locations were translated to units of degrees by multiplying them by 360/256 (i.e., 37 × 360/256 = 52.0). Angles between the horizontal meridian and the superior/inferior peak locations are denoted by α (superior) and β (inferior) in circle 1 and α′ (superior) and β′ (inferior) in circle 2. As the calculation circle is moved temporally (circle 2), the angular degree between the superior and inferior peaks increases: α (52.0°) → α′ (67.5°), β (60.5°) → β′ (66.1°).

The angles between the horizontal meridian and the superior–inferior peak locations were denoted by α (superior) and β (inferior) in circle 1, and α′ (superior) and β′ (inferior) in circle 2. Figure 2C shows that temporal displacement of the calculation circle can widen the peak distance. As the calculation circle is moved temporally (circle 2), the angle between the superior and inferior peaks increases (α<α′, β<β′). The mean angles between the temporal horizontal meridian and the superior/inferior peak locations were 70.6 ± 17.5 and 64.0 ± 10.0, respectively, in circle 1. In circle 2, the mean angles from the temporal horizontal meridian were significantly (P < 0.001) increased in both the superior (77.2 ± 18.7) and inferior peaks (73.3 ± 22.8; Table 6). We presumed that these angle increases were responsible for the decrease in false-positive errors for circle 2. 

Table 6.

 

Mean Angles between the Temporal Horizontal Meridian and the Superior/Inferior Peak Locations

	Circle 1*	Circle 2†	P ‡
Superior peak	70.6 ± 17.5	77.2 ± 18.7	<0.001
Inferior peak	64.0 ± 10.0	73.3 ± 22.8	<0.001

 Data are expressed as means ± SD.

*  Calculation circle automatically placed by the Cirrus™ HD-OCT built-in algorithm.

†  Manually positioned calculation circle centered on the margin involving the temporal edge of the peripapillary atrophy and nasal scleral lip, which corresponds to the neural canal opening.

‡  Paired t-test.

Our results support this hypothesis. Although there were no statistically significant differences regarding the whole-scan thickness, the RNFL thicknesses of circle 2 were significantly thicker in nontemporal sectors and thinner in temporal sectors than those of circle 1. Furthermore, the clock-hour thicknesses were in close agreement with the above-mentioned results. Around the temporal area (i.e., the 7–11 o'clock sectors), the RNFL was significantly thinner in circle 2 than in circle 1. In contrast, around the nasal area (i.e., the 1, 2, 5, 6, and 12 o'clock sectors), the RNFL was thicker in circle 2 than in circle 1 (Table 3). In accordance with these results, the probability of abnormal OCT parameters at the 5% level (i.e., the false-positive rate) was significantly lower in the nasal quadrant and in the 1-, 2-, 5-, and 6-o'clock sectors (Table 4). 

In this study, two different peripapillary RNFL thickness measurements were taken after the location of circle 2 was independently determined by us. The RNFL values of circle 2 were compared with those of circle 1 after we intentionally selected one of our two independent measurements of circle 2 on the basis of the finding that determining the location of circle 2 is highly reproducible (Table 2). Unlike the peripapillary RNFL thickness value, it is impossible to average the color-coded classification, which has a nominal variance, when comparing the probability of abnormal OCT parameters at the 5% level for circle 1. This is why we chose one of the two measurements for circle 2 instead of taking an average. 

We assessed false-positive detection at the 5% level instead of the 1% level because we are of the opinion that the 1% level is too strict an index to use when screening for glaucomatous change in eyes with tilted discs. In addition, the cutoff value of 5% is sufficiently strict for determining abnormality on OCT. It has been reported that RNFL thickness measurements using OCT have high specificity but moderate sensitivity with regard to their normative database. ^17,25,26 If <1% criterion is used to detect RNFL defects, the sensitivity of OCT will be poor. Many studies have used 5% as the cutoff to detect RNFL defects and have reported that this cutoff provided the best diagnostic accuracy on OCT. ^1,25 Therefore, we only used the 5% cutoff level as an index for false-positive detection of abnormal peripapillary RNFL thickness, even though we found that the mean number of abnormal clock-hour sectors at the 1% level was significantly lower in circle 2 (0.49± 0 .94) than in circle 1 (0.33 ± 0.76; P = 0.027), and the proportion of eyes that had one or more abnormal clock-hour sectors at the 1% level in circle 2 (15/69, 21.7%) was lower than that in circle 1 (20/69, 29.0%; P = 0.1797). 

There are some limitations to the present study. First, it is possible that manual displacement of the calculation circle was unintentionally affected by the examiner and would be difficult to reproduce. To minimize the examiner-related bias and low reproducibility, two of the authors shifted the calculation circle in a blinded fashion. Furthermore, we also tried to center the calculation circle on the margin involving the temporal edge of the PPA and nasal scleral lip as accurately as possible on magnified Cirrus HD-OCT fundus images. Although the placement of the calculation circle was highly reproducible between the two authors (Table 2), further investigation is needed to clarify the correct placement of the calculation circle in a more reproducible and standardized manner. 

Second, there were still false-positive results up to 42.0% (Table 5), even after the temporal shift of the calculation circle. This finding implies that factors other than the nasal deviation of the calculation circle likely affect peripapillary RNFL measurement in myopic tilted discs. As the optic disc becomes tilted, the temporal peripapillary ocular tissues stretch and atrophy in myopic eyes. ²⁷ These changes can lead to differences in reflectivity or backscatter detected by the OCT and subsequent differences in the RNFL thickness measurements. In addition, increased axial elongation in myopes may lead to mechanical stretching and thinning of the choroid and retinal pigment epithelium. ²⁸ These changes may affect the ability of the Cirrus HD-OCT inbuilt algorithm to measure RNFL thickness and can cause errors in the RNFL measurements. Furthermore, because the actual scanning radius in eyes with greater axial length (myopic eyes) could be longer than 1.7 mm due to the effect of ocular magnification, measured RNFL thickness can decrease as axial length increases. ^3,6 Although further evaluation is needed, we suspect that the above-mentioned factors can cause errors in the RNFL measurement and could be responsible for false-positive errors in myopic tilted disc. 

Third, our study sample was composed entirely of young male Korean subjects. Such homogeneity may limit the generalizability of our results to other populations with different age, sex, and/or race. However, this homogeneity reduces the potential confounding factors that could affect the thickness of the peripapillary RNFL measured by OCT. 

In summary, we demonstrated that peripapillary RNFL thickness measurement using a manually positioned OCT calculation circle based on the contours of the NCO corresponds more closely to the normative database of the Cirrus HD-OCT than using an automatically determined calculation circle. Although it may not be convenient to apply this technique in clinical practice, this study suggests that the accurate registration of OCT scan positions should be considered when clinicians measure the RNFL thickness in subjects with myopic tilted disc. 

Footnotes

 Disclosure: J.K. Chung, None; Y.C. Yoo, None