Assessment of macular function in patients with non-vascularized pigment epithelial detachment

Non-vascularized pigment epithelial detachments (PED) are usually associated with dry age-related macular degeneration (AMD). In this study, we aimed to investigate the correlation between visual function and morphologic parameters. Seventeen eyes of eleven patients with non-vascularized AMD were enrolled. In addition to conventional optical coherence tomography (OCT), polarization-sensitive optical coherence tomography (PS-OCT) measurements were performed by evaluating the regularity of retinal pigment epithelium (RPE) entropy within the PED area. Retinal sensitivity was measured with MP-3 microperimetry, and retinal sensitivities within (RSin) and outside (RSout) the PED area were calculated. The relationship between OCT parameters and visual function was analyzed. As a result, there was a significant difference between the RSin and RSout (p < 0.001, Wilcoxon signed rank test). Moreover, RSin was significantly related to logMAR VA (p = 0.033, linear mixed model). The regularity of RPE entropy was significantly related to visual acuity and RSin (p = 0.00038, p = 0.031, linear mixed model), although neither the height nor area of PED correlated with visual function. Our results suggest that retinal sensitivity is significantly deteriorated within the PED area and RPE entropy measured with PS-OCT was closely related to visual function in eyes with non-vascularized PED.

www.nature.com/scientificreports/ RPE evaluated by PS-OCT and visual function was investigated in eyes with non-vascularized PED; those with vascularized PED were excluded in order to directly evaluate the effect of changes in the polarization of RPE on retinal sensitivity, minimizing the effect of polarization changes in other retinal tissue. Moreover, in neovascular AMD, retinal sensitivity deteriorates not only due to PED, but also due to other factors such as choroidal neovascularization (CNV), macular edema, and subretinal fluid 26 .
In previous studies, the assessment of visual function in patients with AMD has been predominantly conducted through visual acuity (VA) evaluation which mainly reflects the function at the fovea. Recent reports have suggested the benefit of measuring visual function in a wider macular area using microperimetry, such as MP-3 microperimetry (Nidek, Japan) 27,28 . In the current study, visual function assessment was carried out using VA and the MP-3 microperimetry.

Results
Seventeen eyes of eleven patients (7 men and 4 women) were retrospectively reviewed in the current study. Table 1 shows the baseline characteristics of these patients. Seven eyes had drusenoid PED and ten eyes had serous PED. The mean age (± standard deviation) was 69.7 ± 7.1 years. The values of the PED height (PEDh) and the PED area (PEDa) were 347.2 ± 171.3 um and 10.2 ± 7.5 mm 2 , respectively. The signal of the regularity of RPE entropy (RPEe) was evaluated as "continuous" in 8 eyes and "discontinuous" in 9 eyes.
The retinal sensitivity was 19.0 ± 6.0 dB as an average of the whole 12°. Retinal sensitivity inside the PED area (RSin) was 15.7 ± 6.6 dB, and retinal sensitivity outside the PED area (RSout) was 23.2 ± 4.6 dB, respectively. There was a significant difference between RSin and RSout (p < 0.001, Wilcoxon signed rank test). RSin was significantly related to logMAR VA (Fig. 1, p = 0.033, linear mixed model). Univariate analysis between logMAR VA and the values of age, central macular thickness (CMT), PEDa, PEDh, and RPEe suggested that the RPEe was significantly correlated with logMAR VA (p = 0.00038, linear mixed model). There was no significant correlation between logMAR VA and the remaining variables (p > 0.05). Using the Akaike information criterion (AICc) model selection, only the RPEe was selected in the optimal model, indicating that a continuous RPEe signal was associated with better VA (    Table 3, linear mixed model, AICc = 105.5). There were significant differences in logMAR VA and RSin between continuous and discontinuous RPEe within the PED region ( Fig. 2A, p = 0.00038; Fig. 2B, p = 0.031; linear mixed model).
In subgroup analysis, there were no significant differences in age, logMAR VA and RSin between drusenoid and serous PED (p = 0.99, p = 0.99, p = 0.89, respectively, Wilcoxon rank sum test). Furthermore, there was a significant difference in PEDh between two groups (p = 0.033) but no significant difference was observed in PEDa (p = 0.15). In serous PED subgroup, PEDh was significantly associated with both logMAR VA and RSin (p = 0.030, p = 0.0088, linear mixed model). On the other hand, there was no significant correlation between PEDh and visual function in drusenoid PED.

Discussion
In the current study, we investigated the relationship between morphologic parameters and visual function in patients with non-vascularized PED. As a result, retinal sensitivity inside the PED was significantly deteriorated compared with that outside the PED. Moreover, we found that RPE entropy measured with PS-OCT was closely correlated with both retinal sensitivity measured with MP-3 and logMAR VA.
Our present results suggested that the height and the area of non-vascularized PED were not related to retinal sensitivity and visual acuity; these findings are inconsistent with a previous report 25 . In the previous study, macular function in 18 eyes with drusenoid PED was analyzed using MP-1 microperimetry. As a result, the height and area of the drusenoid PED were significantly associated with retinal sensitivity within the central 4 and 8°. One possible reason for this discrepancy is the difference in patients' background because our present study included eyes with serous PED in addition to drusenoid PED. However, our current result suggested that neither PEDh nor PEDa was associated with visual functions in drusenoid PED subgroup. It might be presumably due to the small sample size in this study and further research is needed to clarify the correlation between OCT parameters and visual function in eyes with non-vascularized PEDs.
Compared to the conventional OCT findings, our result suggested that PS-OCT was useful to predict visual functions in eyes with non-vascularized PED. This may be because PS-OCT measurement sharply reflects the RPE function by detecting RPE melanocytes, whereas conventional OCT can measure only morphological abnormalities such as RPE aperture or intraretinal RPE migration and not any functional deterioration. Furthermore, it is advantageous to use depolarized images even when simply clarifying morphological evaluation of RPE. Miura et al. previously reported that PS-OCT is superior to conventional OCT in detecting focal RPE defect in the serous PED with AMD 23 , supporting this idea. www.nature.com/scientificreports/ Drusenoid PEDs sometimes develop geographic atrophy, which is the main cause of progressive visual loss in dry AMD. In the Age-Related Eye Disease Study (AREDS) report, 19% of eyes with drusenoid PED progressed to geographic atrophy by 5 years follow-up 29 . It is possible that the decreased depolarization (discontinuous RPE entropy) might sharply detect RPE atrophy at the PED. We recently reported that there was a decrease in depolarization at the PED around the RPE aperture, which was mainly found in drusenoid PED eyes 30 . Retinal sensitivity was significantly decreased in the area surrounding the RPE aperture and was associated with the degree of depolarization, demonstrating the presence of RPE atrophy in the decreased depolarization area. PS-OCT provides tissue-specific 3-dimensional information, therefore it might be superior to other ophthalmic examinations in estimating RPE atrophy in eyes with nonvascular PED. Actually, some patients with nonvascularized PED demonstrated no obvious sign of RPE atrophy in fluorescein angiography (FA), indocyanine green angiography (ICGA) and fundus autofluorescence (FAF) but the discontinuous RPE entropy in PS-OCT (Fig. 3).
The present study has some limitations including its retrospective nature and small sample size. In addition, this study was cross-sectional, and, therefore, the change in PED over time cannot be assessed. Our present results suggested that visual function was more closely related to the RPE change evaluated by PS-OCT in eyes with AMD and non-vascularized PED. However, we could not determine whether the change in RPE entropy in patients with PED precedes other ophthalmic findings, such as fundus autofluorescence. Finally, quantitative analysis of RPE entropy was not performed in the current study. It would be of interest to examine whether the quantification of RPE entropy enables to predict visual function in the future. www.nature.com/scientificreports/ In conclusion, the decreased visual function in AMD eyes with non-vascularized PED was significantly correlated with the discontinuous RPE entropy. These findings indicate the usefulness of PS-OCT in evaluating macular function associated with structural changes in non-vascularized PED.

Methods
The Research Ethics Committee of the Graduate School of Medicine and Faculty of Medicine at The University of Tokyo approved this single-center, cross-sectional study. The study protocol adhered to the tenets of the Declaration of Helsinki. Written informed consent was obtained from each patient.
We reviewed the clinical records of patients who were diagnosed with non-vascularized AMD with associated PED at the University of Tokyo Hospital. The following information was apparent from this evaluation: All patients underwent comprehensive ophthalmic examinations including visual acuity and OCT assessment. Spectral domain OCT (Spectralis, Heidelberg Engineering) was used to measure the PEDa, PEDh, and CMT. In eyes with multiple PEDs, the PEDh was measured in the largest PED and the PEDa was calculated as the sum of all PED areas. CNV was not detected by FA, ICGA, FAF and OCTA measurements in all studied eyes. Moreover, no signs of other retinal diseases, such as CSC, were observed during follow-up period.
Retinal sensitivity was measured using fundus-monitored microperimetry (MP-3, Nidek, Japan). A 4-2-staircase strategy with Goldmann III-sized stimuli was used with 25 stimulus locations within 12°, as previously described (Fig. 4A) 31 . To investigate the correlation between visual function and OCT parameters, retinal sensitivity was superimposed on OCT images for all subjects (Fig. 4B). Retinal sensitivities inside (RSin) and outside (RSout) the PED area were calculated. www.nature.com/scientificreports/ PS-OCT measurements were performed using a clinical prototype for retinal imaging (Tomey Corp, Nagoya, Japan). The system of PS-OCT is based on the technology of swept source OCT and the light source was a wavelength-swept laser with a center wavelength of 1050 nm. The optical interferometer was also used to detect the elements of Jones matrix, which mathematically characterizes the polarization property of the target. The area of 6 × 6 mm was scanned using 512 A-lines × 512 B scans in 3.3 s. By averaging the intensities of the Jonesmatrix elements, polarization independent intensity image was obtained from each eye. Polarization scrambling or depolarization was parameterized as entropy. To investigate the regularity of RPE entropy (RPEe), the depolarization images measured with PS-OCT were assessed in all examined eyes (Fig. 5). Using the horizontal scan PS-OCT images, we investigated whether the depolarization signal was continuous along with RPE throughout the PED area for each eye. Two examiners (MK and AF) evaluated the RPEe in eyes with non-vascularized PED. The RPEe was assessed as "continuous" or "discontinuous". In cases where the second examiner did not agree with the first examiner, a panel discussion (MK, AF and TI) was held to draw a conclusion.
The correlation between visual functions (visual acuity [VA] and RSin) and the OCT parameters (age, PEDh, PEDa, CMT, and RPEe) was analyzed using univariate and multivariate linear regression. In addition, using AICc model selection, we investigated which parameter was the best explanatory variable for visual functions. In multivariate regression models, the degrees of freedom decreases with an increasing number of variables; hence, it is recommended to use model selection methods to improve the model fit by removing redundant variables rather than by performing simple multivariate regression analysis, particularly when the number of explanatory variables is large 32,33 . The AIC is an established statistical measure used to evaluate the relationship between variables, and the AICc denotes the corrected AIC, providing an accurate estimation even when the sample size is small 34 . Thus, the optimal model for logMAR VA or RSin was obtained from 2 5 patterns with five variables (age, PEDh, PEDa, CMT, and RPEe). All statistical analyses were performed using the statistical programming language R (R version 3.1.3; The Foundation for Statistical Computing, Vienna, Austria).

Summary statement.
This study aimed to investigate the correlation between visual function and morphologic parameters, including polarization-sensitive optical coherence tomography. Our results suggest that retinal sensitivity significantly deteriorated within the pigment epithelial detachment area. Furthermore, retinal pigment epithelium entropy was closely related to visual function in eyes with non-vascularized pigment epithelial detachments.