Abstract
Purpose.:
To study the effect of contact lens (CL) power on retinal nerve fiber layer (RNFL) thickness measurements using optical coherence tomography (OCT).
Methods.:
Cross-sectional study on 15 healthy subjects. Peripapillary RNFL thickness was measured using time-domain OCT without CL and with CL of different powers (+4 D, +10, −4D, and −10 D), and again without CL. RNFL thickness measurements were compared statistically.
Results.:
Age of study subjects was 30.5 ± 3.2 (mean ± SD) years. Global RNFL thickness measurements showed good repeatability (average without CL was 107 ± 9.3 μm for the first measurement, and 107.4 ± 7.4 μm for the second [P = 0.8]). There was no significant effect of CL power on global RNFL thickness, although more variability was observed with the −10 D CL. When analyzing individual clock-hour segments, only differences in the 6 o'clock hour segment were statistically significant (between −10 D CL [155.9 ± 22.4 μm] and without CL [143.5 ± 19.3 μm], as well as between −10 D and +4 D CL [138.9 ± 19.9 μm], with P < 0.001). When analyzed for individual quadrants, differences in the inferior quadrant were significant between measurements with −10 D CL and all other CLs.
Conclusions.:
Refractive error changes at the corneal plane do not significantly affect RNFL thickness measurements using OCT; caution is warranted in inferior segments, where variations may occur. These results may be applicable to RNFL thickness measurements before and after cataract or refractive surgery.
Optical coherence tomography (OCT) allows for noninvasive imaging of ocular structures. It is widely used in clinical and scientific ophthalmology to obtain high-resolution cross-sections of the retina. Time-domain (TD) OCT uses interferometry, in which low-coherence infrared light is reflected by tissue layers and interacts with light reflected by the reference mirror to produce an A-line signal. Multiple A-line signals are processed to create a high-resolution two-dimensional image of for example the retina, resembling a histologic cross-section.
1,2 The OCT used for this study (Stratus OCT; Carl Zeiss Meditec, Dublin, CA), a TD-OCT, uses a diode light source, emitting light with a wavelength of 820 nm. It has a theoretical axial resolution of <10 μm.
3
Using OCT, the thickness of the retinal nerve fiber layer (RNFL) can be measured. The RNFL may be affected in various diseases; for example, it becomes thinner in glaucoma and optic atrophy, whereas it is thicker in papilledema. Studies have explored factors that can affect RNFL thickness measurements using OCT. In a study on normal children, Salchow et al.
4 found that hyperopic eyes tended to have a thicker RNFL compared to myopic eyes. The same correlation was found in a large population-based study in Australia.
5 When RNFL thickness is measured using OCT, a circular scan is centered on the optic disc. While magnification is not likely to affect the OCT scan in the
z-axis (the A-line measuring thickness of the retinal layers), the diameter of the circular scan has been shown to vary depending on axial length.
6 The same study showed that a larger scan diameter (measuring RNFL thickness further away from the optic disc) results in a thinner measurement.
It is therefore possible that the correlation between RNFL thickness and refractive error, which has been found in some studies, at least partially reflects a systematic measuring error rather than anatomic properties of the eye. This would be of clinical as well as scientific significance, as OCT measurements may be affected by a change in refractive error of the examined eye. If this were the case, measurements before and after refractive surgery, cataract surgery, or with and without contact lenses would not be comparable.
To study the relationship between refractive error and RNFL thickness as measured by OCT, one has to alter the refractive error of a given eye and measure RNFL thickness. In this study, we investigated the effect of contact lens power on RNFL thickness measurements using OCT.
Study subjects were recruited among employees of the Yale Eye Center; participation was voluntary. No subject had a history of ocular disease except refractive error. All subjects received an autorefraction or manifest refraction. An external and anterior segment examination was unremarkable in all cases.
One randomly selected eye of each subject was used for study purposes. The cornea was anesthetized with one drop of proparacaine hydrochloride 1% and the pupil was dilated with phenylephrine 2.5% eye drops to a diameter of at least 5 mm. Measurements of peripapillary RNFL thickness were obtained with a TD-OCT system (Stratus OCT, software version 4.0.1; Carl Zeiss Meditec) using the fast RNFL scan mode. The photographer (AMH or DJS) centered each scan on the optic disc with the aid of a light fixated by the subject and a video image showing the scan in relation to the optic disc. Three circular scans (diameter of 3.4 mm), each comprised of 256 A-lines, were acquired. The RNFL was automatically identified by the OCT based on its reflectivity. The photographer reviewed all scans for centration, quality of focus, and completeness of data. Inadequate scans were discarded and the measurement was repeated until satisfactory scan quality was achieved. Three measurements were averaged for analysis. The 360° of the circular scan were subdivided into quadrants and clock-hour segments. The average thickness of the RNFL over 360° is referred to as global RNFL thickness.
4,7 Refractive error and axial length were not entered into the OCT. The focus on the OCT was adjusted to obtain maximal signal strength and to allow the subject to view the fixation light clearly. Only scans with signal strength of ≥ 6 were accepted. OCT data were processed with the software provided by the manufacturer (Carl Zeiss Meditec) for the fast RNFL scan (RNFL Thickness Average Analysis Report 4.0.1). The final printout was used for statistical analysis.
RNFL thickness was first measured without CL. Soft CLs (Proclear Sphere, base curvature 8.6 mm, diameter 14.2 mm, center thickness 0.065 mm for −3 D CL per manufacturer; CooperVision Inc., Fairport, NY) of different powers were then worn in random order during measurements. Proper position and centration of the CL in the eye was verified. Four different CL powers (−10 D, −4 D, +4 D, +10 D) were used to assess the effect of a 20 D change in refractive error on RNFL thickness measurements. Finally, RNFL thickness was again measured without CL to assess repeatability. The two measurements without CL were averaged and the average was used for comparison with other measurements.
Statistical analysis was performed using commercial software (SAS 9.2; SAS Institute, Inc., Cary, NC). For purposes of statistical analysis, the clock-hours of the left eye were mirror-imaged so that 3 o'clock was nasal and 9 o'clock was temporal for all eyes. A mixed model repeated measures analysis was used to compare RNFL thickness across CL powers. This analysis appropriately accounted for the correlation among the repeated measures within the same eye. Post-hoc pairwise comparisons adjusted by Tukey's method were used to maintain overall type I error at 0.05 level, which was considered statistically significant unless otherwise indicated.
For purposes of this study, a difference between RNFL thickness measurements was considered clinically significant if it was greater than 10 μm. This threshold was chosen, because the intervisit SD has been shown to be 2.5 μm,
8 thus two standard deviations on either side will equal 10 μm.
The study was approved by the Yale University Human Investigation Committee and is in compliance with the Human Insurance Portability and Accountability Act (HIPAA) regulations, as well as the tenets of the Declaration of Helsinki. Written informed consent to participate in this study was obtained from all subjects.
Fifteen healthy subjects participated in this study, average age was 30.5 ± 3.2 (mean ± SD) years; the range was 25 to 36 years. There were 10 male and 5 female subjects; 6 were Caucasian, 8 were Asian, and one subject was Hispanic. Mean refractive error (spherical equivalent) was −3.9 ± 4.7 D (range −14.75 to +3.5 D). Eight right eyes and 7 left eyes were studied.
Average global RNFL thickness without CL was 107.2 ± 8.2 μm (mean ± SD). There was good repeatability of global RNFL thickness without CL (first measurement was 107 ± 9.3 μm, second measurement was 107.4 ± 7.4 μm [
P = 0.8]). RNFL thickness in different quadrants without CL is shown in
Table 1. Overall there were no significant differences between two measurements without CL. In eight subjects, the difference between the first and second global RNFL thickness measurement without CL was <5 μm, and in 14 it was <10 μm (in one subject, the difference was 12 μm).
Figure 1 shows the repeatability of RNFL thickness measurements as a function of location. Average signal strength difference was statistically significant between scans obtained with −10 D CL and no CL, but not in any other comparison.
Results for global RNFL thickness measured with different CLs are shown in
Figure 2. No significant effect of CL power on global RNFL thickness measurements was observed, but there was more variation (larger SD and range) with the −10 D CL compared to all other CL powers. When comparing global RNFL thickness measurements with different CLs, we found a statistically significant difference between the −10 D CL and the −4 CL (
Fig. 3) although the 95% CI for this difference did not include the threshold for clinical significance. All other comparisons did not yield clinically or statistically significant differences.
Results of RNFL thickness measurements stratified by clock-hour segments are shown in
Table 2. The only statistically significant differences were found in the 6 o'clock segment, between measurements with −10 D CL versus without CL, and between −10 D CL versus +4 D CL (a more stringent criterion of
P < 0.001 for significance of pairwise comparison was chosen to account for multiple comparisons made for individual clock-hour segments).
Table 3 shows RNFL thickness measurements stratified by quadrants. Only differences in the inferior quadrant reached statistical significance.
The interaction between CL and quadrant was statistically significant (
P = 0.03), indicating that the magnitude of the differences between CLs was dependent on which quadrant was measured. Differences between CLs were significant only in the inferior quadrant, but not in others. Post hoc pairwise comparisons for the inferior quadrant are shown in
Figure 4. Differences between measurements with −10 D CL and all other CLs were significant. The strongest effect was observed when comparing RNFL thickness measured with −10 D CL and with +10 D CL.
The 95% CI for differences between measurements also extended beyond 10 μm (clinical significance) for these comparisons: −4 D CL versus +4 D CL, −4 D CL versus +10 D CL, and no CL versus +10 D CL.
In this study, we investigated whether changing the refractive error of an eye at the corneal plane has an effect on RNFL thickness measurement using OCT. The refractive error was varied by using CLs with different powers. This is similar to changes encountered after and refractive surgery and possibly after cataract surgery, or if a patient wears CLs for one measurement but not for the next. The CL power was varied by 20 D, which is clinically a large range. To assess repeatability, we performed two separate measurements without CL, one in the beginning and one at the end of the session. There was very good agreement between the two measurements (in only one subject was the difference for global RNFL thickness greater than 10 μm, the threshold for clinical significance in this study).
Appropriate OCT signal strength is necessary to achieve valid and reliable RNFL measurements.
9 All scans met our minimum criterion (signal strength = 6), therefore signal strength is unlikely to have significantly affected our results. Although the difference in signal strength was significant betweens scans obtained with −10 D CL and no CL, no other comparisons showed significant differences.
Several studies have found that RNFL thickness measured with OCT is affected by the refractive error of the eye. Salchow et al.
4 calculated from their cohort of normal children that an increase in spherical equivalent by 1 diopter was associated with an increase in RNFL thickness of 1.7 μm. A similar effect was found in a large population-based study on children in Australia.
5 Both studies are cross-sectional in design; although they suggest an effect of refractive error on RNFL thickness, the refractive error was not varied in a given eye to investigate its effect on OCT measurements of RNFL thickness.
Longitudinal measurements on eyes after changes in refractive error have been published. Sharma et al.
10 measured RNFL thickness using OCT in eyes with mild to moderate myopia before and after LASIK and LASEK. Mean average (global) RNFL thickness was 98.2 ± 5.6 μm before and 98 ± 6.01 μm after LASIK, and 98.5 ± 5.9 μm before and 99.1 ± 6.2 μm after LASEK; the differences were minor and statistically insignificant. Another study on RNFL thickness after LASIK supported these findings.
11 El-Ashry et al.
12 measured RNFL thickness in patients before and after cataract surgery. They found that after cataract surgery, the RNFL was thicker by 8.5 ± 8.2 μm, which was statistically significant. The largest differences were noted in the nasal followed by the inferior quadrant. The authors attributed the differences to a better scan quality after cataract surgery. Since our subjects were young adults and did not have any disturbances of the refractive media, this potential source of error does not apply to them.
Bayraktar, Bayraktar, and Yilmaz
6 found that the diameter of the circular OCT scan, used to measure peripapillary RNFL thickness, was affected by axial length but not by refractive error. Although it is held that magnification does not affect the axial scans (in the
z-axis), it will likely affect the horizontal (
x- and
y-axes) dimension of an OCT scan, in our case the diameter of the circular scan. In fact, Bayraktar et al.
6 found that the radius of the circular scan, which was set to 1.73 mm, actually varied from 1.51 to 1.87 mm. On average, the difference between the preset value and the actual radius was small (0.05 ± 0.09 mm), but the range was significant. For each 1 mm increase in axial length, the actual projected scan radius increased by approximately 0.06 mm or 3.5%. Furthermore, when the examiner adjusted the scan radius to 1.73 mm, global RNFL thickness was significantly different compared to the unadjusted scan. When no manual adjustment of the actual projected scan radius was done, thinner RNFL thickness measurements were found for longer eyes, and thicker measurements for shorter eyes. This was attributed to differences in scan radius, since larger radii yielded thinner RNFL thickness measurements, compared to smaller radii.
6
To correct for potential effects of magnification on OCT measurements of the RNFL, El-Dairi and coworkers
13 adjusted their measurements for axial length. There are several ways to calculate magnification of ocular fundus structures,
14 but these are rarely used in clinical practice. For this reason, we did not enter axial length or refractive error into the OCT machine, and only adjusted the focus to ensure that the fixation light could be seen clearly by the subject, and that the scan circle was centered well on the optic nerve. To overcome the effect of magnification when comparing scans in different persons, the cross-sectional area of ganglion cell axons converging onto the optic disc may be calculated by multiplying the magnification-corrected scan circumference with the RNFL average thickness.
5,6 Since our goal was to investigate the effect that changing the refractive error of a given eye at the corneal plane has on OCT measurements of RNFL thickness, we did not incorporate axial length into our measurements and analysis.
Youm et al.
15 found that global RNFL was thicker when measured with CL compared to without (their study did not examine the effect of CL power on the measurement, but rather asked whether wearing a CL itself affects the measurement). Even though these differences reached statistical significance, they were insignificant in absolute numbers (99.4 ± 9.7 μm with CL versus 100.8 ± 10.3 μm without CL in CL wearers, and 102.8 ± 10.8 μm with CL versus 105.3 ± 9.9 μm without CL in non-CL wearers.
15 In our study, global RNFL measurements did not consistently demonstrate this pattern; measurements with −10 D CL yielded thicker RNFL measurements than those without a CL and with +10 D CL. Moreover, there was no consistent trend in our results toward thicker or thinner RNFL with positive or negative CL power.
A large study on normal children found the greatest variation of RNFL thickness in the 12 o'clock and 6 o'clock sectors.
5 In a series on healthy 12-year old children, Wang et al.
16 also found reproducibility to be lowest in the inferior quadrant. These factors may partially explain our observation that RNFL thickness measurement with −10 D CL were significantly different in the inferior quadrant, but not in other quadrants. We did observe that fixation was difficult for the study subjects when wearing the −10 D CL. This could have led to decentration of the scan and further contributed to the differences observed in the inferior quadrant. Finally, CL centration on the cornea may have been less optimal with the thicker −10 D CL compared to thinner CLs, causing optical aberrations.
Huynh et al.
5 found that RNFL thickness decreased with increasing axial length, but increased with more positive spherical equivalent. Rauscher et al.
17 found that moderately myopic subjects tended to have relatively thin peripapillary RNFL when measured with an OCT (Stratus OCT; Carl Zeiss Meditec). This was most pronounced in the superior and inferior quadrants. In that study, global RNFL thickness decreased by 7 μm for every mm of axial length, and by 3 μm for every diopter of myopia. Both effects were statistically significant, though the effect of axial length on RNFL thickness was stronger. It therefore appears that anatomic properties influence OCT measurements of RNFL thickness more than optical properties of the eye. Our findings support this, as we did not find CL power to affect RNFL thickness measurements using OCT in a systematic way.
In clinical practice, changing the refractive error of an eye by 20 D represents a major change. This would be encountered if, for example, a patient with pathologic myopia wore a CL for one OCT measurement but not for the next. After cataract extraction and intraocular lens implantation or after refractive surgery, differences of this magnitude are rarely observed. Finally, one should keep in mind that our results do not apply to differences in refractive error caused by axial length.
We interpret our results that refractive error (changed at the corneal plane with contact lenses) does not significantly affect RNFL thickness measurements using OCT. Caution is warranted in the inferior quadrant, where variations may occur. For clinical practice, RNFL thickness measurements with and without CL should be comparable, and this may also apply to measurements before and after cataract or refractive surgery.
Supported in part by a departmental Challenge Grant from Research to Prevent Blindness, Inc., New York, NY, to the Department of Ophthalmology and Visual Science, Yale University School of Medicine.