Abstract
Purpose:
The purpose of this study was to identify the relationship between aniseikonia scores in the vertical and horizontal meridians and the foveal microstructure on vertical and horizontal spectral-domain optical coherence tomography (SD-OCT) in patients with idiopathic epiretinal membrane (ERM).
Methods:
All patients (n = 65) with unilateral ERM were examined, and the aniseikonia scores in the vertical (VAS) and horizontal (HAS) meridians were determined using the New Aniseikonia Test. Vertical and horizontal images passing through the fovea were obtained by axial SD-OCT in both eyes. The thicknesses of the ganglion cell layer + inner plexiform layer, inner nuclear layer (INL), and outer retinal layer were measured on the SD-OCT images, and color histograms were analyzed using Photoshop software.
Results:
Of the 65 ERM patients, 81.5% (53 patients) had macropsia. The VAS and HAS were equal in 52.8% (28 patients). Multiple regression analysis revealed significant correlations between the VAS and vertical INL thickness (R = 0.388, P = 0.001) and between the HAS and horizontal INL thickness (R = 0.349, P = 0.001). The difference between VAS and HAS was proportional to the ratio of the vertical INL thickness to horizontal INL thicknesses (R = 0.370, P < 0.001).
Conclusions:
Eyes with ERM mostly presented macropsia. The aniseikonia scores in the vertical and horizontal meridians correlate well with INL thickness on the vertical and horizontal directions of SD-OCT images, respectively. Aniseikonia induced by ERM may be related to the INL thickening detected with SD-OCT.
An epiretinal membrane (ERM) is a nonvascular fibrocellular proliferation on the retinal surface. The incidence of spontaneous ERM is 5.3% in patients over 49 years of age.
1 Although previous study samples have been small, 78–100% of ERM patients reportedly have aniseikonia.
2–6 Aniseikonia is an ocular condition characterized by a significant difference between the eyes in the perceived size of images. ERM-induced aniseikonia is presumably due to compression or stretching forces that cause an abnormal distribution of the retinal receptors,
3 creating a perceived image that is larger or smaller in size. Recently, adaptive optics-scanning laser ophthalmoscopy (AO-SLO) of eyes with ERM showed microfolds in the foveal photoreceptor layer, suggesting shrinkage of the ERM might cause contraction in the photoreceptor layer, thus producing microfolds.
7 Patients with ERM exhibit more severe aniseikonia and are at greater risk of binocularity loss and developing asthenopia and impaired vision quality.
5
The New Aniseikonia Test (NAT; Handaya, Tokyo, Japan) measures aniseikonia. This computerized test compares the sizes of red and green semicircles between the eyes by triggering dissociation with red/green glasses.
8 The size of each eye's semicircle is adjusted until the shape is perceived as a perfect circle. The NAT presents two series of matched semicircles (vertical and horizontal) at random, and the aniseikonia scores are measured in the vertical and horizontal meridians. A recent study used the NAT to examine the relationship between the aniseikonia scores and retinal microstructures in patients with ERM.
6 However, only one report investigated the relationship between the aniseikonia scores and retinal microstructures, and the study only used the mean aniseikonia scores. In contrast, numerous studies have examined the relationship between metamorphopsia and ERM.
9–14 Furthermore, these studies measured the differing metamorphopsia between horizontal and vertical lines.
9–12 Metamorphopsia, which is a symptom described as perceived distortion of objects, is different from aniseikonia. However, both aniseikonia and metamorphopsia have a similar development mechanism, which reflects abnormal distribution of photoreceptors.
6 There is usually metamorphopsia, together with aniseikonia from retinal distortion caused by the ERM. Reportedly, the horizontal metamorphopsia scores increase according to ERM severity but do not change with the vertical metamorphopsia scores, and the horizontal metamorphopsia scores are significantly larger than the vertical metamorphopsia scores in advanced stages of ERM.
9 However, there are no known studies evaluating the correlation between the vertical meridian and horizontal meridian aniseikonia scores and the ERM microstructure.
Spectral-domain optical coherence tomography (SD-OCT) is commonly used to assess microstructural changes in patients with macular disease, and the relationship between visual function and OCT findings has received great attention in patients with ERM.
15 Therefore, we investigated the difference in the aniseikonia scores between the horizontal and vertical meridians and the relationship between aniseikonia and ERM microstructures.
We analyzed a total of 65 patients with idiopathic unilateral ERM at HanGil Eye Hospital from September 2014 to March 2015. This retrospective case series was performed in accordance with the Declaration of Helsinki and was approved by the institutional review board. The 65 ERM patients were diagnosed using clinical findings from the ophthalmic examination and SD-OCT (Spectralis; Heidelberg Engineering, Vista, CA, USA). Patients with any of the following conditions were excluded: evidence of ocular inflammation, diabetic retinopathy, age-related macular degeneration, retinal vasculitis, history of intraocular surgery other than uncomplicated cataract surgery, anisometropia greater than 2.0 diopter (D), and a logarithm of the minimal angle resolution (logMAR) best-corrected visual acuity (BCVA) > 1.0. We also excluded eyes for which accurate OCT images could not be obtained. To compare both eyes, the contralateral eye was also examined.
Ophthalmic examination included the BCVA, slit-lamp biomicroscopy, indirect ophthalmoscopy, and the NAT. The BCVA was measured with a Snellen chart, and the decimal visual acuity was converted to the logMAR. Manifested refraction was measured as sphere, cylinder power, and cylinder axis. To analyze astigmatism, we converted the refractive astigmatism from the spherocylinder notation to power vector notation by applying a Fourier transformation as follows
16:
J0 = −
C/2 × cos2α, where
C is negative cylindrical power and α is the cylindrical axis.
J0 refers to cylinder power set at orthogonal 90 and 180 meridians, representing Cartesian astigmatism. This notation allows a direct comparison to be made of the subject's astigmatism separately for the horizontal and vertical meridians.
17
Aniseikonia severity was assessed in all eyes by using a computerized NAT composed of matched pairs of red/green semicircles with a white round fixation target on a black background; binocular vision was dissociated using red/green goggles. The fixation target measured 4 cm in diameter (visual field angle ∼5.7°), and the red semicircle measured 40 cm in diameter. The diameter of the green semicircle varied in 1% increments from 1% to 24%. The subjects viewed the monitor from a 66-cm distance with the appropriate correction. Each subject was instructed to identify which semicircle pair appeared equal in size, and the size difference in the semicircle pair represented the percentage of aniseikonia. Two series of matched semicircles were presented at random, and the aniseikonia scores were calculated for the vertical meridians (VAS) and horizontal meridians (HAS). The difference between the meridians, defined as the DAS, was calculated by subtracting the VAS from the HAS. Aniseikonia of 2% or greater was considered macropsia and aniseikonia of −2% or less as micropsia.
SD-OCT comprising 25 single horizontal axial scans (scanning area: 6 × 6 mm, centered at the fovea) was performed in both eyes to assess the central retinal structures. Two cross-sectional retinal layer images were selected in the vertical and horizontal directions for each eye. The vertical OCT image provided a cross section of the retinal layer vertically through the fovea, and the horizontal OCT image showed a cross section of the retina horizontally through the fovea. The central macular thickness (CMT) and total macular volume (TMV) were calculated automatically from the OCT data by three-dimensional averaging using the OCT software. The retinal layer thickness was measured using Photoshop software (Adobe Systems, San Jose, CA, USA). Each retinal layer of the SD-OCT axial scan image was moved using the Photoshop dragging tool, and a color histogram of each dragged image was examined. The pixel value of each dragged area in the individual frames was calculated and defined as the thickness of each retinal layer (
Fig. 1). The thicknesses of the retinal layers were measured in both the horizontal and vertical directions as follows: ganglion cell layer (GCL) + inner plexiform layer (IPL); inner nuclear layer (INL) thickness; and outer retinal layer (from the outer nuclear layer to the cone outer segment tips). Because we experienced considerable intraobserver variability when measuring the thickness of the GCL, IPL, outer plexiform layer, outer nuclear layer, and photoreceptor layer separately, the thickness of the GCL + IPL and outer plexiform layer to RPE was measured as a single unit. The thickness of each retinal layer was determined by agreement between two masked, well-trained observers (H.C. and G.S.), and the mean was used for analysis. The ratio of INL thickness was calculated by dividing the vertical INL thickness by the horizontal INL thickness. The ratios of the GCL + IPL and outer retinal layer thickness were calculated in the same manner. The mean (SD) was calculated for each OCT parameter and the aniseikonia scores.
The difference in the OCT parameters between the ERM and contralateral eyes was analyzed with the Mann-Whitney U test. Univariate and stepwise multiple linear regression analyses were used to assess the relationship between possible influencing factors and vertical or horizontal aniseikonia scores. All variables with P < 0.10 in the univariate analysis were included in the multiple regression model. Multicollinearity was evaluated among predictor variables by using the variance inflation factor. We used variance inflation factor >10 as a guide for exploring alternative models. The associations between the DAS and the ratio of the GCL + IPL, INL, and outer retinal layer thickness were evaluated using the Spearman rank correlation test. All tests of associations were considered statistically significant at P < 0.05. The analyses were performed using SPSS version 18.0 for Windows (SPSS, Inc., Chicago, IL, USA).
The patient characteristics are summarized in
Table 1. Of the 65 ERM patients, 53 (81.5%) exhibited macropsia (range, 2%–19%). Twelve patients did not exhibit aniseikonia. Micropsia was not observed in any patients.
Mean VAS and HAS were 5.41 ± 5.41% (range, 0%–19%) and 4.89 ± 4.83% (range, 0%–19%), respectively (
Table 1). The VAS was slightly larger than the HAS, but the difference was not significant. Twenty-five (47.2%) patients showed a 2% or greater difference between the VAS and HAS. Ten (18.9%) patients exhibited a VAS that was less than 2% smaller than the HAS. Fifteen (28.3%) patients had a VAS that was more than 2% greater than the HAS.
There was no significant correlation between the BCVA and aniseikonia scores. The aniseikonia score was not correlated with the spherical equivalent or the difference in the spherical equivalent between both eyes. There was no correlation between the vectorial notation of refractive astigmatism and aniseikonia scores (
Table 2).
The VAS correlated with the vertical GCL + IPL thickness (
R = 0.241,
P = 0.015), vertical INL thickness (
R = 0.333,
P = 0.001), horizontal GCL + IPL thickness (
R = 0.227,
P = 0.024), and horizontal INL thickness (
R = 0.231,
P = 0.021) on univariate analysis. The HAS also correlated with the vertical GCL + IPL thickness (
R = 0.217,
P = 0.030), vertical INL thickness (
R = 0.235,
P = 0.017), horizontal GCL + IPL thickness (
R = 0.193,
P = 0.046), and horizontal INL thickness (
R = 0.312,
P = 0.002) on univariate analysis. However, the aniseikonia score did not correlate with the vertical and horizontal outer retinal layer thicknesses (
Table 2). Multiple linear regression analysis was performed using the vertical GCL + IPL thickness, vertical INL thickness, horizontal GCL + IPL thickness, and horizontal INL thickness as predictors. A significant positive correlation was seen between the VAS and vertical INL thickness (
R = 0.388,
P = 0.001) on multiple linear regression analysis, but no correlation was observed between the VAS and horizontal INL thickness. A positive correlation was also seen between the HAS and horizontal INL thickness (
R = 0.349,
P = 0.001) on multiple linear regression analysis, but not between HAS and vertical INL thickness. Our multicollinearity check did not reveal any problems, with all predictor variables in the multivariate models having a variance inflation factor < 3.
The DAS was proportional to the ratio of INL thickness (
R = 0.370,
P < 0.001). However, the ratio of the GCL + IPL was weakly correlated (
R = 0.181,
P = 0.058), and outer retinal layer thickness did not correlate with the DAS (
R = −0.176,
P = 0.083;
Table 3;
Fig. 2).
We found that 53 of 65 patients (81.5%) with ERM had aniseikonia, with all 53 patients exhibiting macropsia. The prevalence of ERM-induced aniseikonia is unknown because it is not routinely tested in a clinic. ERM is a relatively common disease, and a high proportion (78%–100%) of ERM patients reportedly have aniseikonia according to a few reports.
2–6 The aniseikonia scores ranged from 2% to 19% in the present study, which is consistent with findings in previous reports.
2–6 The ERM with macular contraction is considered a representative macropsia disorder. Presumably, the compression caused by shrinkage of ERM results in a more packed distribution of retinal receptors; as a result, incoming light stimulates more receptors, and the image appears larger.
3,7
The VAS (5.41 ± 5.41%) was slightly larger than the HAS (4.89 ± 4.83%) in the present report, although this difference was not significant. These results are consistent with the findings of previous reports.
6 Aniseikonia measured by computerized NAT was considered reproducible and reliable according to previous reports.
3,18 Validation studies of computerized NAT revealed 0.990 ± 0.005 horizontal and 0.991 ± 0.004 vertical correlation coefficients and 0.985 ± 0.111 horizontal and 0.989 ± 0.102 vertical slope, suggesting a small underestimation.
3 Yoshida et al.
18 also reported that the aniseikonia measured with the NAT was 1.4% less than when measured with the phase difference haploscope, with no significant difference in the horizontal and vertical directions. Previous reports suggest that the small NAT underestimation is clinically insignificant, and there is quite good agreement in the vertical and horizontal directions. Therefore, we suggest the NAT used in this study is a reliable method for measuring aniseikonia. We could not calibrate the NAT personally, however, lending a level of uncertainty to our NAT-determined aniseikonia score.
It is well known that anisometropia causes aniseikonia.
19,20 More than 2.0 D of anisometropia was excluded from the study to minimize anisometropia-induced aniseikonia. We believe aniseikonia was not related to the spherical equivalent or the difference of spherical equivalent between both eyes. To account for patients' astigmatism, we applied the vectorial notation of spherical-cylindrical refraction, but found no correlation between aniseikonia scores and astigmatism.
Mean INL thickness has been associated with mean aniseikonia scores.
6 We also observed an association between vertical and horizontal INL thickness and vertical and horizontal aniseikonia scores, respectively. The consistency between this and a previous study suggests that the INL change may be one of the important etiologies of aniseikonia. We hypothesized how morphologic alterations mostly involving the INL could displace the structure of photoreceptors, leading to aniseikonia. First, aniseikonia might be caused by a spherical lensing effect of microcystic changes in the INL. Microcystic changes, predominantly involving the INL, may result from a mechanical stretch threshold for which cell loss occurs and are a common SD-OCT finding in ERM.
21,22 In our study, 39 of the 53 ERM patients with macropsia (73.6%) exhibited microcystic changes in the INL on SD-OCT (
Fig. 3). In addition, the AO-SLO imaging revealed extra microcystic changes not seen on SD-OCT.
23 Cynthia et al.
23 observed the cones underlying the microcystic change appeared more tightly packed in the AO-SLO and suggested paraxial optical ray tracing modeling the microcystic change as a spherical lens. Based on this modeling, we postulated the light rays bending of the path caused by the cyst might stimulate more photoreceptors. Thus, the magnification lens effect of INL microcystic change might lead to macropsia. Moreover, aniseikonia is known to be not reduced after surgery, although the INL thickness was significantly decreased than preoperative values.
6 INL microcystic changes are also observed after membrane removal.
21 Therefore, this postulation might explain why aniseikonia did not change after surgery. Second, the correlation between the aniseikonia scores and INL thickness may be coincidental. The INL is the location of many early inflammatory and tractional pathologic spaces.
24 In the process of transmission of contraction force of an ERM to the photoreceptor layer, INL thickening may occur, but abnormal distribution of photoreceptor cells might eventually occur, resulting in aniseikonia. Because current SD-OCT image does not represent photoreceptor disarrangement, we could not observe the correlation between the aniseikonia scores and photoreceptor disarrangement; thus, the observed relationship between the aniseikonia scores and INL thickness may have been artifactual. Previous evidence supports our suggestion that disarrangement of photoreceptor cells correlates with the severity of metamorphopsia using AO-SLO in eyes with ERM.
7
We found that more severe vertical INL thickening correlated with an increased severity of vertical aniseikonia, and more severe horizontal INL thickening correlated with an increased severity of horizontal aniseikonia. However, these results contradict those of previous studies examining the relationship between ERM and metamorphopsia. One study found that horizontal metamorphopsia in ERM was correlated with the vertical retinal contraction, and vertical metamorphopsia was correlated with the horizontal retinal contraction. The authors surmised that this correlation may be caused by the directionality of retinal plasticity. The optic disc may restrict horizontal displacement of the posterior retina, and as a result, the vertical plasticity of the posterior retina may be greater than that of the horizontal retina. If the sensory retina contracts vertically, then the photoreceptors would also be dislocated vertically, and eventually, the distortion of the horizontal line would increase (i.e., an increase in horizontal metamorphopsia).
10,11 Metamorphopsia is a type of distorted vision in which a grid of straight lines appears wavy, whereas aniseikonia is an ocular condition in which the perceived size of images differs significantly. Thus, if the sensory retina is contracted in the vertical direction, the photoreceptor layer would also be compressed vertically, eventually increasing the vertical macropsia. Therefore, we speculated that when the retinal structures contracted vertically, the INL would exhibit greater thickening on vertical OCT, and eventually, the VAS would increase.
The DAS was proportional to the ratio of INL thickness (
Table 3;
Fig. 2). Thus, as the difference between the VAS and HAS increases, wider variations are observed in the INL thickness on vertical and horizontal OCT. The NAT is used to compare the perceived size of semicircle images between both eyes; however, it is easier to compare the length (top and bottom edges) of the semicircle meridian than the actual size of the semicircle. Theoretically, the vertical NAT is compared with the vertical semicircle size, but in practice, the vertical NAT is compared to the length of the vertical semicircle meridian. Comparison with the length of each vertical semicircular diameter may affect the vertical meridian passing through the fovea of the retina. In the present study, the stimulated vertical meridian of the retina was measured on the vertical OCT (
Fig. 4). Therefore, we presumed that each VAS and HAS was more closely associated with the INL thickness on the vertical and horizontal OCT images, and the DAS was proportional to the ratio of INL thickness. If the difference between the VAS and HAS is larger, then the difference in INL thickness on vertical and horizontal OCT may be greater (
Fig. 5).
The difference between vertical and horizontal INL was 7944 − 7869 = 75 pixels, which translates to approximately 9.4 μm. The difference seemed small, but the average difference in INL thickness as measured by two manual graders is 2.22 ± 1.30 pixels,
25 where the mean difference between the actual thickness (Spectralis OCT) and predicted thickness ranged from 0.65 to 1.01 μm.
26 A prior report demonstrated that changes in INL thickness before and 6 months after ERM removal were a small but significant 7.6 μm.
14 Thus, we suggest the difference between vertical and horizontal INL (75 pixels) is meaningful with regard to the accuracy of OCT measurements.
In clinical practice, we recommend the measurement of aniseikonia in ERM patients on initial examination. We advise careful measurement of INL thickness on vertical and horizontal OCT if there is a substantial observed difference in aniseikonia scores in the vertical and horizontal meridian.
The limitations of this study include a relatively small sample size, measurements based on single horizontal and vertical cross sections centered at the fovea, and manual measurement using Photoshop software. In addition, we could not calibrate the NAT personally, lending a level of uncertainty to our NAT-determined aniseikonia score. Future studies with a large sample size that include multiple cross-sectional images and automated computerized measurements are needed.
In conclusion, this study revealed that most patients with ERM have macropsia. The VAS and HAS were associated with the vertical and horizontal INL thicknesses, respectively, whereas the DAS was proportional to the ratio of INL thickness. This study may facilitate future investigations of the relationship between aniseikonia and retinal microstructures and in the interpretation of aniseikonia test results.
Disclosure: H. Chung, None; G. Son, None; D.J. Hwang, None; K. Lee, None; Y. Park, None; J. Sohn, None