Relative Ellipsoid Zone Reflectivity in Macular Telangiectasia Type 2

Purpose The relative ellipsoid zone reflectivity (rEZR) has been proposed as an innovative biomarker for photoreceptor integrity. This study evaluates the rEZR in macular telangiectasia type 2 (MacTel) eyes of different disease stages. Methods The mean rEZR (ratio ellipsoid zone [EZ]/external limiting membrane [ELM] reflectivity [arbitrary units {AUs}], grey level range = 0-1) was analyzed for an entire spectral domain optical coherence tomography volume scan (global) and for each subfield of the Early Treatment Diabetic Retinopathy Study (ETDRS) grid (topographic) in patients with MacTel and controls. MacTel disease severity was classified according to Gass and Blodi. Results Linear mixed-model analysis of 145 eyes of 74 patients and 50 eyes of 25 controls revealed globally lower, yet not statistically significant, rEZR values in MacTel eyes. Topographically, most pronounced decreases were found in stages 3 and 4/5 for the temporal inner (coefficient estimates [CEs] = −25.4 [−38.2; −12.6] and −34.1 [−48.7; −19.6] AU, both: P < 0.001), the inferior inner (−29.9 [−44.6; −15.6] and −35.3 [−52.1; −18.5] AU, both: P < 0.001), the nasal inner (−21.5 [−35.52; −7.4] and −31.6 [−47.6; −15.6] AU, P = 0,003 and P < 0.001), and in the superior inner subfield of stage 4/5 (−25.0 [−42.0; −7.9] AU, P = 0.004). Conclusions The rEZR showed association with disease severity and the predilection area of MacTel. Given the current understanding of the pathophysiological concept of MacTel, these findings underscore the value of the rEZR as a potential novel biomarker for outer retinal integrity. Longitudinal studies are demanded to better characterize its value as a biomarker for early photoreceptor alterations and disease progression in MacTel.

M acular telangiectasia type 2 (MacTel) is a bilateral disease of the central retina, typically affecting individuals over 40 years at first onset. 1,2 Structural alterations, for example, proliferation of intraretinal vessels and loss of the ellipsoid zone (EZ) occur initially in the temporal parafovea, which later involve an oval shaped area of 8 degrees horizontal and 5 degrees vertical diameter centered on the fovea ("MacTel area"). 2 Disease progression also leads to photoreceptor dysfunction resulting in loss of retinal sensitivity and finally visual acuity. [3][4][5] In spectral-domain optical coherence tomography (SD-OCT), photoreceptor loss in MacTel is detectable as a slowly progressive disruption of the second hyper-reflective band in the outer retina, termed the EZ. 2 Although its quantification has become an accepted outcome measure in ongoing phase III trials (NCT trial number NCT03319849 and NCT03316300), biomarkers and treatment opportunities targeting already earlier in the disease course would be desirable. Herein, identification of high-risk eyes and/or macular regions prior to the occurrence of irreversible photoreceptor alterations and functional impairment would be needed, but reliable biomarkers are still missing. [6][7][8][9][10][11] Given the fact that retinal visualization by SD-OCT imaging is based on refractive tissue properties, photoreceptors' mitochondria are currently assumed to represent the source of the EZ's signal. In addition, functional complexes between Müller cells and retinal photoreceptors become apparent in SD-OCT imaging as the first hyper-reflective band and is named the external limiting membrane (ELM). The ratio of the EZ to the ELM reflectivity, termed the relative EZ reflectivity (rEZR), has recently been proposed as an auspicious biomarker for outer retinal integrity and photoreceptor health. 9,[12][13][14] First studies revealed an association of the rEZR with disease severity in age-related macular degeneration (AMD). 15,16 In MacTel, however, only the pure (and uncorrected) EZ reflectivity has been assessed by Barthelmes et al. showing reduced reflectivity values in the parafoveal region. 17 Although the study by Barthelmes and colleagues was performed in only 14 patients with MacTel, the results warrant a refined characterization of reflectivity changes in the outer retina in eyes with MacTel.
The purpose of this study is to assess the rEZR within the pathophysiological concept of MacTel with regard to disease staging and its spatial predilection of structural alterations in the retina as well as to evaluate the rEZR in MacTel eyes without manifest EZ loss. 2 The evaluation of the rEZR within the pathophysiological concept of MacTel will be of informative value regarding its usefulness as a biomarker for outer retina integrity and photoreceptor degeneration.

MATERIALS AND METHODS
For this cross-sectional case-control prospective study, patients with MacTel and controls were selected from a single center cohort at the Department of Ophthalmology, University of Bonn, Germany. Patients were included between October 2016 and November 2022 in the context of the Natural History and Observational Registry study (NHOR; www.mactelresearch.org), a multinational natural history study on MacTel. Inclusion criteria were a confirmed MacTel diagnosis and retinal imaging of sufficient quality. Exclusion criteria included the presence of confounding ocular conditions, such as AMD and diabetic retinopathy, as well as optic media opacities. Controls were identified after a comprehensive ophthalmic examination. If both eyes of one participant fulfilled the inclusion criteria, both eyes were included. This study was approved by the local ethics committee at the University Hospital Bonn (Bonn, Germany) and adhered to the tenets of the Declaration of Helsinki and written informed consent was obtained prior to data acquisition.
All participants underwent a comprehensive ophthalmologic examination, including the assessment of bestcorrected visual acuity (BCVA), a slit-lamp and dilated fundus examination, and retinal imaging. Following pupil dilatation with tropicamide 0.5% and phenylephrine 2.5% eye drops, retinal imaging was performed including 55 degrees color fundus photography (CFP; Zeiss Visucam 500; Carl Zeiss Meditec, Dublin, CA, USA), combined confocal scanning laser ophthalmoscopy (cSLO; near-infrared reflectance [IR] 30 degrees, automated real-time mode [ART] at least 30 single frames) and SD-OCT (30 degrees × 25 degrees, 121 B-scans, high speed mode; Spectralis HRA2+OCT; Heidelberg Engineering, Heidelberg, Germany). The 30 degrees × 25 degrees grid was used to allow for rEZR assessment within and beyond the MacTel area.

Image Preprocessing and Grading
MacTel disease stages were classified according to Gass and Blodi by CFP assessment. 18 Gradings were performed by two masked readers (authors L.G. and S.K.). In case of disagreement, a third reader (author S.T.) was involved to arbitrate. Manifest EZ loss was determined in SD-OCT Bscans and defined as the complete absence of the EZ. Mere attenuations of the signal with the EZ still being present were not judged as manifest EZ loss. 9 Delineation of retinal areas with manifest EZ loss was performed in the transverse SD-OCT image presentation of the built-in Heidelberg Eye Explorer software (version 2.5.5; Heidelberg Engineering)  yellow lines within the white rectangle) for exemplary determination of the relative ellipsoid zone reflectivity (rEZR). Further, each ROI is presented magnified for the logarithmic and linear displayed SD-OCT image. Corresponding reflectivity profiles for each ROI show the specific peaks for the external limiting membrane (ELM; also as the magnified picture), ellipsoid zone (EZ), interdigitation zone (IZ), and the retinal pigment epithelium (RPE). Please note, the overall outcome of this study is the mean rEZR, which was determined as the ratio of the EZ to ELM reflectivity for the entire volume scan (global assessment) and for each of the Early Treatment Diabetic Retinopathy Study (ETDRS) subfields (topographical assessment). For the purpose of better representation, only a single reflectivity profile of a representative ROI is here presented for controls (A) and each MacTel disease stage (B-F). generating en face image slabs. 10 The so-called "photoreceptor 1" line, overlaying the EZ, was used, and manually corrected if needed, as the reference and the distance to a second line was set to 0, hence the thickness of the line was set to 1 pixel. Areas of EZ loss were delineated using the "draw region" tool and, if indicated, its boundaries adjusted after re-assessment in the traditional B-scan view. Figure 1 depicts an exemplary case for the described method. Any area of EZ loss was excluded from further statistical analysis.

Determination of the rEZR
The rEZR, defined as the EZ/ELM reflectivity ratio, was determined, as previously described, in reflectivity profiles in the raw SD-OCT image files using a semi-automated approach using MatLab (version 9.5; The MathWorks, Natick, MA, USA; annotated code available at: https://github.com/ bisselma/relEZIquantification). 19 The ELM's reflectivity has been postulated to be stable across a wide eccentricity, to be present in the fovea, and to be one of the retinal layers undergoing least reflectivity alterations with age in most diseases. 15,20 Although Müller cell alterations are known to occur in pathophysiological process of MacTel, these are described to be limited mainly to areas with already existing EZ loss. 4,21 As these areas were systematically excluded from statistical analysis in the here presented study, the ELM was maintained as the reference for the rEZR calculation.
Further, automated retinal layer segmentation was performed, and, if required, manually corrected using the Heidelberg Eye Explorer software (HEYEX, software version 1.10.4.0; Heidelberg Engineering). Segmentation coordinates, exported as XML files, were superimposed to the OCT raw images (i.e. native, non-logarithmic transformed data) and used for straightening of each OCT B-scan along with the coordinates of the Bruch's membrane enabling accurate rEZR determination even in eyes with pronounced posterior pole curvature. Within every single OCT B-scan, the rEZR data was obtained at adjoining regions of interests (ROIs) in corresponding reflectivity profiles (dynamic range of grey values: 0-1 [arbitrary units {AUs}]), see Figure 2A. The width of each ROI was set at 10 pixels (approximately 120 μm in high-speed SD-OCT imaging) along the image x-axis.

Global and Topographic rEZR Assessment
To assess the rEZR both globally (i.e. within the entire SD-OCT volume scan) and topographically (i.e. focusing the "MacTel area"), the Early Treatment Diabetic Retinopathy Study (ETDRS) grid was used to determine the mean rEZR within each of the nine ETDRS subfields. 22 Figure 3 gives a detailed representation of the position of the "MacTel area" within the ETDRS grid.

Statistical Analysis
Patient characteristics are presented as absolute and relative frequencies for categorical variables and as means with standard deviations (SDs) for continuous variables. To investigate the primary research question of differences (after exclusion of retinal areas with manifest EZ loss) between rEZR values in MacTel and control eyes, linear mixed-effects models were used. These accounted for possible correlations between measurements taken from the same patient and/or eye (nested data structure with multiple data points within one eye, and 2 eyes per patient). Differences in the mean rEZR (arbitrary unit [AU]) between MacTel and control eyes were assessed (i) for the total SD-OCT volume raster scan (global assessment) and (ii) for the individual subfields (topographic assessment) of the ETDRS grid. Additionally, in eyes with MacTel, we considered retinal disease stages according to Gass and Blodi to investigate the association between MacTel disease severity and rEZR. Linear mixedeffect models were adjusted for age and eccentricity, entering the model as a spline (B-spline of degree 2). The P values <0.005 were considered significant (due to adjustments for multiple testing by Bonferroni-correction for 10 subfields). In order to better understand the cohort's data structure, a refined analysis of the study cohort was further performed and aimed in evaluation of the inter-eye correlation of cases and controls with bilateral inclusion as well as of the heterogeneity of the data set. The inter-eye correlation was assessed in determining Spearman's correlation coefficient r, which was calculated for the mean rEZR of the total SD-OCT volume raster scan of each participant with both eyes included. Further, heterogeneity of the data was evaluating by comparing the variance as well as the range of the rEZR values between patients with MacTel and controls.
For a refined characterization of the rEZR as a potential biomarker, we conducted an exploratory subgroup analysis of MacTel eyes which did not show manifest EZ loss. Here, P values <0.05 were considered significant. The analyses were performed with the software R, version 4.1.2, using the package lme4. 23   Descriptive analysis revealed a median (first quartile and third quartile) rEZR per SD-OCT volume scan of 36.9 (11.6 and 79.2) AU in MacTel eyes and of 56.9 (17.7 and 111.1) AU in control eyes. Regarding the refined analysis of the cohort's data structure, Spearman's correlation coefficient was assessed for the mean rEZR per volume raster scan in all 25 included controls and in 71 out of 75 patients with MacTel. It was revealed to be r = 0.164 for controls and r = 0.134 for patients with MacTel indicating a minor intereye correlation. The variance of rEZR values for MacTel eyes was more than five times higher than that of controls (24.15 vs. 135.61 AU). Consistent with these findings, the range of rEZR values in MacTel eyes was wider (on average 9.14 AU) than in controls (on average 8.40 AU). Accordingly, MacTel eyes tend to have a higher degree of heterogeneity and more extreme values than controls. Bilateral data were available from every control patient and from 71 out of 74 patients with MacTel.

Cohort Characteristics
Given the inhomogeneous sample size across MacTel disease stages, results of MacTel diseases stage 1 and stage 2 will be reported in the following as one group (MacTel stages 1/2) as well as of stage 4 and stage 5 (MacTel stages 4/5). Across MacTel disease stages, the median rEZR per SD-OCT volume scan was revealed to be 39.5 (15.4 and 82.9) AU in MacTel stages 1/2 eyes (n = 58, 36 patients),  Table 1 gives an overview of the cohort characteristics.

Global rEZR Assessment Across MacTel Disease Stages
Linear-mixed models, accounting for patients' age and the topographic dependence of the 36,626 ± 9109 (mean ± SD) rEZR data points within the SD-OCT raster scan, were applied for refined analysis of the rEZR across different MacTel disease stages. Compared to controls, global assessment revealed a decreased rEZR of all MacTel eyes  Figure 2 shows exemplary reflectivity profiles for distinct ROIs for each disease stage.

Topographic rEZR Assessment
The rEZR decrease of MacTel eyes showed a topographic dependency being more pronounced in the direct parafovea compared to the foveal or more eccentric retinal region. Biggest rEZR differences between controls and MacTel eyes across all disease stages were found in the temporal inner (P < 0.001), inferior inner (P < 0.001), and nasal inner (P = 0.004) ETDRS subfield with CEs of −23.  Table 2.
The associations of MacTel disease staging and the retinal region with the rEZR remained apparent when considering both simultaneously. The mean rEZR of each ETDRS subfield was shown to be more decreased with advancing disease stages and this decrease, again, was more pronounced in the pericentral ETDRS subfields. Compared to healthy eyes, the rEZR in the temporal inner subfield was decreased with a CE of  Table 3. For a graphical presentation of the results, see Figure 4.
Linear-mixed models revealed that all included MacTel eyes without manifest EZ loss exhibited a decreased rEZR with a CE of −5.  Table 4 summarizes the topographical results across the total MacTel subgroup without EZ loss in detail.
With regard to disease staging of the included MacTel eyes without manifest EZ loss, topographic analysis revealed the rEZR decrease to be more pronounced with advancing disease staging in all of the four inner ETDRS subfields. Herein, assessed stage 3 MacTel eyes without manifest EZ loss, exhibited a statistically significant rEZR decrease in the temporal inner and the inferior inner ETDRS subfield indicating CEs of −21.  For detailed results of the topographic rEZR analysis in MacTel eyes (without manifest EZ loss) of different disease stages, please see Table 5. Figure 5 demonstrates a graphical representation of rEZR results in MacTel eyes without manifest EZ loss for both global and topographical assessment.

DISCUSSION
This is the first study evaluating the rEZR, a potential novel biomarker for outer retinal integrity, in patients with MacTel, demonstrating lower rEZR values in MacTel eyes compared to controls. This difference was revealed to underlie a spatial dependence with pronounced differences in the predilection area of MacTel which, moreover, were also apparent in eyes without manifest EZ loss. Additionally, the rEZR was shown to be associated with disease severity indicated by lower rEZR values in advanced MacTel stages.
Given current assumptions of photoreceptors' mitochondria and their junctional complexes with Müller cells being the source of the EZ and ELM reflectivity signal in SD-OCT imaging, respectively, their assessment has become a well-established indicator for photoreceptors' integrity and outer retina degeneration in MacTel. 24,25 Whereas a qualitative, structural loss of EZ and ELM indicates not only fulminant and progressive, but also and more importantly irreversible damage of the outer retina, a quantification of the EZ and ELM reflectivity signal in SD-OCT imaging might be helpful here to assess earlier outer retinal changes preceding and/or occurring beyond manifest EZ and ELM loss.
This study demonstrates in the topographical assessment that across all included MacTel eyes the rEZR decrease was most and significantly pronounced in the temporal inner and inferior inner ETDRS subfield with a CE of −23.2 (−35.7 and −10.8) AU and of −25.2 (−39.4 and −11.1) AU (both: P < 0.001). In the global analysis, however, the rEZR decrease in MacTel eyes was revealed to be both less pronounced (CE of −7.7 [−17.8 and 2.5] AU) and not statistically significant (P = 0.141). These findings are in line with previous studies highlighting the topographical predilection of retinal alterations in MacTel being mostly located temporal and inferior to the fovea, in the so-called "MacTel area." 2,26-28 Barthelmes et al. have previously identified an EZ reflectivity reduction in MacTel eyes, which was also spatially confined to the temporal-inferior region of the parafovea. 17 Given the chosen study design by Barthelmes and colleagues, including the small sample size of 14 MacTel eyes, the use of a different (and no longer available) OCT device as well as having not normalized the EZ reflectivity (and thus not avoiding potential localized confounders on the EZ signal), their results need to be interpreted very cautiously. But nevertheless, the findings by Barthelmes et al. and our here presented findings indicate underlying photoreceptor/outer retina affection in the perifoveal region, notable also in eyes without manifest EZ loss, which is topographically in line with the current understanding of MacTel pathophysiology. 17 Besides the topographical aspect, one further important finding of this study is the rEZR's association with MacTel disease severity. In the global and the topographical analysis, the rEZR was shown to decrease with advancing disease stages of MacTel (see Table 3, Fig. 4). Interestingly, we found the inter eye correlation to be reduced in eyes of more advanced disease stages. Although distinct MacTel-associated structural alterations have been identified in SD-OCT imaging, only limited data are available on their precise association with disease staging according to Gass and Blodi. [29][30][31] The here presented study, however, shows an association of the rEZR with the CFPbased disease staging, which again highlights underlying and in SD-OCT imaging detectable photoreceptor and outer retina alterations present already in early disease stages and which increase given the chronic-progressive nature of MacTel.
Several limitations need to be considered in the interpretation of this study. Although this is the first study evaluating the rEZR in MacTel, other concurrent SD-OCT based features of MacTel have not been considered here. Second, this study aimed in only assessing structural image data of patients with MacTel and does therefore not include any analysis of visual function and its correlation with the rEZR. Further, because analyses were performed on a cross-sectional image data set, further longitudinal studies are needed to evaluate the rEZR's prognostic value for disease progression. However, a strength of this study is the prospectively acquired image data set in context of an international multicenter natural history study (www. mactelresearch.org) warranting therefore high quality and comparability. Additionally, it needs to be mentioned that the rEZR was determined as the ratio of the EZ to ELM reflectivity (see Methods section for further details). As explained, the ELM is assumed to represent junctional complexes between Müller cells and photoreceptors, which can both alter during the disease process of MacTel. 32 To date, histopathologic analysis of Müller Cells in MacTel has only been performed in a small number of eyes, large scale analyses are lacking. 4 But whereas ELM alterations on high resolution retinal imaging have been described predominantly in areas of manifest EZ-loss, their impact on the present analysis is here assumed to be negligible as EZ loss areas were systematically excluded from any analysis. 25,[33][34][35] In conclusion, this is the first study assessing the rEZR in MacTel revealing a decrease of the rEZR being associated with disease staging and which was spatially more pronounced in the parafoveal predilection area of MacTel. These findings are in line with the current understanding of the MacTel pathophysiology supporting the rEZR to be a potential innovative SD-OCT biomarker for outer retinal integrity and MacTel severity. This study warrants further analyses, including a refined structure-function-correlation and evaluation of the rEZR's prognostic value for disease progression in MacTel.

FIGURE 5.
Graphical representation of linear-mixed model results from the subgroup analysis, for example, macular telangiectasia (MacTel) eyes without any manifest ellipsoid zone (EZ) loss, for the global (first row) and the topographic (second to final row) assessment of the relative ellipsoid zone reflectivity (rEZR) and its association with MacTel disease staging. The extent of rEZR differences between MacTel stages and healthy controls is indicated by the color-coded (from light red to dark blue) coefficient estimates. Coefficient estimates of higher negative values, indicating more pronounced rEZR decreases of MacTel eyes compared to controls, are represented by darker blue colors. P values (*** = P value < 0.001; ** = P value <0.01; * P value <0.05). Eyes with MacTel disease stages 4/5 were not included in this analysis, as all of them did show manifest EZ loss and did hence not fit into the analyzed subgroup.