Retinal nerve fibre layer (RNFL) thickness measurement by optical coherence tomography (OCT) may be a useful adjunct for early glaucoma diagnosis. The frequently1^–3 used RNFL thickness scan protocols employ a fixed 3.4 mm diameter scan-circle around the optic disc (fig 1 top). This diameter was originally4 arbitrarily decided on, since it was large enough to avoid overlap with the optic nerve head (ONH) in nearly all eyes and yet allow measurement in an area with a thicker RNFL than expected with a 4.5 mm diameter circle. It has recently been shown by ultrahigh-resolution OCT5 that there is least variability in RNFL measurements 1.7 mm from the centre of the disc, which corresponds to the location of the 3.4 mm diameter circle scan used in the Stratus OCT.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Distance from the disc margin varying with scan protocol used. Normal-sized optic disc (A) scanned by 3.4 mm scan circle protocol (B). Large disc (C) scanned with the 3.4 mm scan circle protocol (D). Note the decreased distance from the optic nerve head compared with (A) and (B). Large disc (E) scanned with 2.27×disc proportional scan protocol (F). Note the similarity of distance from the optic disc margin compared with (A) and (B).

Since RNFL thickness decreases with increasing distance from the ONH,6 larger discs would be likely to demonstrate thicker RNFL values than smaller discs simply because of the shorter distance from the edge of the ONH where it is measured (fig 1C,D). The use of the 3.4 mm scan protocol for large discs may thus lead to fallaciously thick RNFL measurements.

The 3.4 mm scan diameter is 2.27 times the size of an average disc measuring approximately 1.5 mm. The Stratus OCT has a proportional scan protocol RNFL thickness (2.27×disc), which acquires a single circle scan around the optic disc, which is 2.27× the radius of the aiming circle. The aiming circle can be adjusted according to the optic disc size. This protocol can be customised to each individual optic disc imaged and enables variations in size of the optic disc to be accounted for. Though the scan diameter would differ in different-sized discs, all scan circles would be at a distance of 2.27× the diameter of that particular optic nerve head (fig 1E,F). This should logically measure the RNFL of all optic discs at comparable distances from the edge of the ONH, regardless of its size.

This study was carried out to compare RNFL measurements using the fast RNFL thickness (3.4 mm) scan protocol and the RNFL thickness (2.27×disc) proportional scan protocol in normal, disc suspect and glaucomatous eyes with different-sized optic nerve heads.

METHODS

This was a prospective cross-sectional study including suspected glaucoma patients and patients with primary open-angle glaucoma (POAG). Normal controls were selected from age- and sex-matched healthy adults. Subjects fulfilling the inclusion criteria detailed below were prospectively recruited for the study. The study was cleared from the Institute Ethics Committee and adhered to the principles enshrined in the Declaration of Helsinki. Informed consent was obtained from all individuals.

Each subject underwent a comprehensive ophthalmic examination including best-corrected visual acuity (BCVA), intraocular pressure (IOP) measured by Goldmann Applanation tonometry, slit-lamp biomicroscopy, gonioscopy and stereoscopic disc assessment using a 90.0 dioptres (D) lens. Colour stereoscopic optic-disc photographs were taken on the Zeiss Fundus camera FF 450 with Visupac System 451 (Carl Zeiss Meditec Ophthalmic Systems, Jena, Germany). Optic discs were assessed by two graders independently, masked to other examination results. All subjects underwent baseline Standard Achromatic Perimetry (SAP) on the Humphrey Field Analyzer HFA 750 II (Carl Zeiss-Humphrey Systems, Dublin, California) using the 24-2 testing protocol by the SITA-Standard strategy. The visual fields were considered satisfactory if false-positive and false-negative errors did not exceed 30%, and fixation errors did not exceed 25%. Any visual-field defect had to be confirmed on a field test done within 1 month from the initial test.

To be eligible for inclusion, normal subjects were required to fulfil the following criteria in both eyes: BCVA ⩾20/40 (±5.0 D spherical; ±3.0 D cylinder); IOP <21 mm Hg; open angles on gonioscopy and normal-appearing optic disc (defined as one with no features suggestive of glaucomatous optic neuropathy such as cup–disc ratio >0.6, any diffuse or focal neuroretinal rim (NRR) thinning, any disc haemorrhage and/or any RNFL defects on the red-free photograph). Patients were included only if both observers classified the disc as “normal.” The visual fields were to be normal, defined as mean deviation (MD) and pattern standard deviation (PSD) values within 95% confidence intervals and a Glaucoma Hemifield Test (GHT) classified as “within normal limits.”

Eyes with any optic-disc finding suggestive of glaucomatous optic neuropathy but with other features of normal subjects were included as glaucoma suspects. POAG patients were included if they had the same BCVA criteria detailed above, open angles on gonioscopy, characteristic glaucomatous optic neuropathy with corresponding reproducible abnormal visual-field tests, defined as an MD and PSD outside 95% normal confidence limits and GHT “outside normal limits” in two consecutive tests.

Patients were excluded if they had any history of ocular disease or intraocular surgery or if they were found to have any ocular disease except glaucoma during examination, such as diabetic retinopathy, uveitis or cataract.

OCT technique

All included subjects were scanned after pupillary dilatation with the Stratus OCT version 4.0.1 (Carl Zeiss Meditec, Dublin, California) by an experienced operator (SK) masked to the patients’ diagnosis.

The OCT machine was aligned. The contralateral eye was covered and internal fixation target used. Once the optic nerve head came into focus, the Z-Offset was adjusted to view the OCT image, and the polarisation was optimised to maximise the reflective signal. All scans were acquired three times during the same day, and an average of the three readings was computed in each case. To be acceptable for inclusion, the signal strength had to have been at least 6, there were to be no missing areas in the scan due to blinks, and the scan must have been properly centred on the optic disc. (The importance of centring was demonstrated by Gabriele et al,7 who showed that even small shifts of the scan circle resulted in significant changes in RNFL measurements.)

The peripapillary RNFL was scanned using the fast RNFL thickness (3.4) scanning protocol and the RNFL thickness (2.27×disc) proportional scanning protocol. For the fast RNFL thickness scan, the operator manually adjusted the aiming circle such that its centre coincided with the centre of the optic disc, ensuring an OCT image of even radius of 1.73 mm around the optic nerve head, regardless of the disc size. For the RNFL thickness 2.27×disc scan, the operator adjusted the aiming circle to approximate the outline of the individual optic disc being scanned as closely as possible. The average RNFL thickness measurements using both scan protocols were recorded.

The disc size was measured using the fast optical disc scanning protocol, which acquires six equally spaced line scans 30° apart running through a common central axis, placed in the centre of the optic disc. The Stratus OCT detects the optic disc margin by automatically detecting the termination of the retinal pigment epithelium (RPE), with a provision of manual correction of this assessment if not satisfactory. All optic-disc scans were checked for the accuracy of the automated detection. Any inaccuracies were manually corrected by the operator and the scans reanalysed using the corrected determination. The disc area was measured by the inbuilt OCT ONH analysis algorithm.

All data were recorded on prospectively completed data forms.

STATISTICAL METHODS

Sample size

Previous reports have demonstrated a difference of 5.8 μm when measuring RNFL thickness in healthy eyes using standard and custom protocols8 of the OCT. The standard deviation of the difference was 10.2 μm.9 For a study with 85% power at 5% significance level, the sample size calculated was (1.96+1.03)²×10.2²/5.8² = 27.64. At least 28 patients would be needed in each category.

Analysis

The results were analysed using the SPSS for Windows, Version 10.0 (SPSS, Chicago). The Wilcoxon signed-rank test was used to analyse differences between RNFL thickness measurements using both scan protocols in the same disc. The Spearman rho correlation coefficient was used to analyse the relationship between optic-disc size and RNFL thickness. In glaucomatous eyes, multivariate analysis using a general linear model (GLM) was used to analyse the effect of disc area and MD (on visual fields) on RNFL measurement using both scan protocols.

RESULTS

Thirty-four normal, 64 suspected glaucoma and 40 glaucoma patients were enrolled for the study between July and December 2006. Two disc-suspect and four glaucoma patients were excluded due to suboptimal OCT scans. Two normal controls were excluded, since they did not return for the visual-field test. Data from 32 normal subjects, 62 glaucoma-suspects and 36 glaucoma patients were analysed.

The refractive error was similar in all three groups (table 1, ANOVA p = 0.48). Glaucoma suspects had significantly larger discs compared with glaucomatous eyes (p = 0.01; Mann–Whitney U test; table 1). RNFL measurements were significantly thinner using the proportional scanning protocol compared with the standard 3.4 mm protocol in all three groups (p<0.001, table 1).

In normal eyes, the optic-disc area did not affect RNFL measurement using the fixed-diameter protocol (Spearman rho correlation coefficient 0.019; p = 0.918; fig 2). Using the proportional scan protocol, the RNFL thickness measurement was inversely proportional to optic-disc area (Spearman rho correlation coefficient −0.638; p<0.001; fig 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Scatter diagram showing no correlation between disc area and retinal nerve fibre layer (RNFL) thickness using a fixed-scan protocol (A) and inverse correlation with customised scan protocol (Y) in normal subjects.

In glaucoma suspects, the optic-disc area correlated significantly with RNFL measurement using the fixed-size scan protocol (Spearman rho correlation coefficient 0.452; p<0.001), but not with the proportional scan protocol (Spearman rho correlation coefficient 0.153; p = 0.234, fig 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Scatter diagram showing a direct correlation between disc area and retinal nerve fibre layer (RNFL) thickness using the fixed-scan protocol (A) and no correlation with customized scan protocol (B) in disc suspects.

In glaucomatous eyes, the mean deviation was −11.9 (SD 6.8) dB. Although the RNFL thickness using the fixed-diameter scan protocol directly correlated with the disc area (Spearman rho correlation coefficient 0.371; p = 0.025; fig 4), on multivariate analysis RNFL measurements were significantly affected by the mean deviation on visual fields, but not by disc area (p<0.001 and p = 0.641 respectively). Using the proportional scan protocol, the RNFL thickness was affected by both the mean deviation and the disc area on multivariate analysis (p<0.001 and p = 0.047 respectively). The mean deviation and disc area in these eyes also correlated significantly (Spearman rho correlation coefficient 0.466; p = 0.004).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Scatter diagram showing no correlation between disc area and retinal nerve fibre layer (RNFL) thickness using the fixed-scan protocol (A) and customised scan protocol (B) in glaucomatous eyes.

DISCUSSION

RNFL thickness measurement by Stratus OCT is fairly well established as a useful adjunct in glaucoma management. Although the spectral-domain OCT is now commercially available and allows greater resolution in imaging retinal layers, its potential use as a tool for glaucoma is yet to be determined. Further, the Stratus OCT is unlikely to be discarded in the face of new technology simply because it contains baseline and follow-up data of thousands of patients worldwide.

Studies have demonstrated a direct correlation between disc size and RNFL thickness8 9^–14 in normal eyes. These data differ from most published reports on the subject. Using the 3.4 mm scan circle in normal subjects, we found no correlation between optic disc size and RNFL thickness. This agrees with Hougaard et al,15 who also found no correlation between optic disc size and RNFL thickness measured by OCT.

This may indicate that RNFL thickness is dependent on the distance from the centre of the optic disc rather than the point of exit from the scleral canal, as hitherto believed. The RNFL would be measured at similar distances from the centre of the disc, regardless of the size of the scleral canal. Blumenthal et al16 also recently found RNFL thickness to be inversely related to the distance from the centre of the optic disc.

Racette et al17 reported larger optic discs and thicker RNFL measurements in healthy eyes of black individuals compared with Caucasians. Savini et al18 commented that this may be related to a measurement artefact on the OCT. In reply, it was rationalised that if a large and small disc had the same number of nerve fibres, the large disc would provide more space for the fibres at the disc margin and so would in fact have thinner RNFL measurements at the disc margin than the smaller disc. At any fixed distance from the centre of the optic disc, mean RNFL-thickness measurements would be the same for optic discs containing the same number of fibres, regardless of the disc size. They hypothesised that the thicker RNFL found in larger discs of African–Americans was due to a higher number of nerve fibres rather than a measurement artefact.

It is difficult to explain the lack of correlation between disc size and RNFL thickness measurements in our study if indeed larger discs contain more nerve fibres. It may be that they are less crowded in large discs and may be spread out more thinly in the peripapillary area. This may offset the presence of a higher number of nerve fibres and may explain why we did not observe thicker RNFL measurements in large discs.

The 2.27×disc scanning protocol yielded thinner RNFL measurements compared with measurements with the fixed-diameter scan circle. The proportional protocol was derived from the theoretical assumption of mean disc diameters of 1.5 mm and the “ideal” 3.4 mm diameter scan circle (3.4/1.5 = 2.27). However, in our study, the mean disc diameters in normal subjects were 2.74 (0.33) mm. When scanning with the 2.27×disc protocol, the actual scan circle would have diameters larger than 3.4 mm, which may have resulted in thinner RNFL measurements because of the greater distance from the ONH.

Glaucomatous eyes showed a direct correlation between RNFL measurements and optic-disc area using the fixed-diameter scan. However, on multivariate analysis, there was no relation to disc size, but the mean deviation significantly affected the RNFL measurements. The mean deviation on the visual fields also correlated significantly with disc area, probably implying that larger discs had less severe glaucoma. We have observed that in our hospital-based referral practice, larger discs with large cups are more likely to be glaucoma suspects, whereas patients with smaller discs and large cups are likely to be glaucomatous. Whether the less severe glaucoma observed in large optic discs is due to earlier referral following earlier detection of large cups or a greater reserve in these eyes owing to a higher number of nerve fibres cannot be commented on from our study.

Interestingly, we found that RNFL measurements correlated to the disc size of glaucoma suspects when using a fixed scan protocol. By definition, their visual fields were normal, so the confounding factor of mean deviation could not be assessed in this group as for glaucoma. However, the possibility of preperimetric glaucoma in small discs cannot be ruled out. Since larger discs had a thicker RNFL, it is possible that small discs with large cups should be followed up more carefully than large discs, where the large cups may reflect only the larger disc area.

Savini et al8 also examined custom scan protocols 0.5 mm and 1.0 mm from the optic nerve head edge. Interestingly, they reported a negative correlation between RNFL thickness and optic-disc size using customized scans. However, if the distance from the optic disc margin was important, then since their customized scans measured all optic discs 0.5 mm and 1.0 mm from the optic-disc margin, the RNFL thickness should not have differed. If the distance from the centre of the optic disc determines RNFL thickness regardless of the size of the scleral canal, then the negative correlation can be easily explained. Even while using a customised scan, larger optic discs would naturally be scanned at a greater distance from the centre of the optic-nerve head.

The three groups were comparable with regard to their refractive correction. We did not use a magnification correction for the refractive error because the range of refractive error included in this study would have had a negligible effect on the size of the scanning circle.19 In a previous study using OCT 2, Bayrakter et al20 studied the effect of magnification of the scan circle radius and reported no correlation between the actual projected scan circle radius and the refractive error.

RNFL thickness probably does depend on the distance from the centre of the disc rather than the disc margin as hitherto believed. Our study indicates that disc size may not matter while measuring RNFL thickness by OCT using fixed-diameter scans. In large discs, proportional scan protocols may yield fallacious results by virtue of measuring RNFL further away from the disc and therefore should be avoided.