PURPOSE
To evaluate the association of hyperreflective foci contiguous with the retinal pigment epithelium (rpeHRF) with visual function impairment in aged normals, early age-related macular degeneration (eAMD), and intermediate AMD (iAMD).
DESIGN
Prospective cohort study.
SUBJECTS
Participants of the MACUSTAR study.
METHODS
MACUSTAR participants underwent color fundus photography, optical coherence tomography (OCT) imaging, best corrected visual acuity (BCVA), low-luminance VA (LLVA), rod-mediated dark adaptation (RMDA) at 12°, contrast sensitivity (CS), mesopic (mesPSD), and scotopic pointwise sensitivity deviation (scPSD) testing. rpeHRF presence and count were determined using custom FiJi software. Group comparisons and associations with visual function were analyzed using analysis of variance, linear regression, and Spearman correlation.
MAIN OUTCOME MEASURES
Presence, burden, and topographic distribution of rpeHRF, as well as association with functional parameters, were determined.
RESULTS
Fifty-six normal aged (33 female, mean age 68.1 ± 6.4 years), 34 eAMD (27 female, 71.7 ± 6.4 years), and 583 iAMD eyes (387 female, 72.0 ± 7.0 years) were included. rpeHRF counts were 0.16 ± 0.85 in normals, 0.33 ± 0.96 in eAMD, and 1.61 ± 2.49 in iAMD ( P <.001). BCVA, LLVA, CS (all P <.001), and scPSD ( P =.001) differed between disease groups, whereas RMDA and mesPSD did not. In iAMD, eyes with rpeHRF showed worse BCVA, LLVA, CS, scPSD (all P <.001), and mesPSD ( P =.02). rpeHRF was found to be associated modestly with only CS, scPSD, LLVA, and BCVA.
CONCLUSIONS
Presence and burden of rpeHRF were independently associated with impaired visual function and may thus serve as a prognostic biomarker for disease progression and enrichment criterion for future interventional trials.
INTRODUCTION
A ge-related macular degeneration (AMD) is the most common cause of legal blindness among the older population in high-income countries. Treatments are well established for the late-stage manifestation of macular neovascularization worldwide ,, and have recently become available (in the United States, Australia, and Japan) for the late stage of geographic atrophy. ,, Dietary supplements delay progression to late-stage disease in fellow eyes of macular neovascularization eyes, and they may slow centripetal atrophy expansion in foveal-sparing geographic atrophy. However, interventions targeting early AMD (eAMD) and intermediate AMD (iAMD), for example, before irreversible vision loss occurs, represent an unmet need.
To enable a realistic clinical trial design in a slowly progressing chronic disease, eyes at high risk for progression need to be identified for enrichment purposes. A variety of prognostic biomarkers visible on high-resolution retinal imaging have been identified in recent years. These include intraretinal hyperreflective foci (iHRF), ,,, which are visible on optical coherence tomography (OCT) B-scans. They are defined as small distinct lesions within the neurosensory retina of similar reflectivity as the retinal pigment epithelium (RPE) with a clear separation from the RPE layer. , In the context of AMD, iHRF are thought to represent RPE cells that leave the RPE monolayer and migrate toward the inner retinal layers. Despite robust supporting histologic evidence, ,, and longitudinal tracking of their formation originating from the RPE, other authors assume that iHRF represent cells of microglial origin. Also, a combination of both is a possible option.
In non-neovascular AMD, iHRF correspond to focal hyperpigmentation on color fundus photography (CFP), ,, a clinical sign associated with progression to late AMD. ,,,, However, hyperpigmentation on CFP also can be related to HRF that are still connected to the RPE, focal thickening of, or an increase in reflectivity of the RPE monolayer itself. ,, These alterations have been termed rpeHRF. Longitudinal tracking using OCT has shown that iHRF are spatiotemporally associated with rpeHRF, indicating that the former arise from the latter. It hence seems intuitive that rpeHRF represent RPE cells at the beginning of their migration toward the inner retina.
Berni and associates recently found that rpeHRF predict the onset of large hypertransmission defects (LHyperTD) independently of drusen volume or iHRF, which suggests that they may serve as biomarkers for iAMD progression to late-stage disease.
Various visual function tests have been evaluated for their predictive value regarding the onset and early progression of AMD. Of these, rod-mediated dark adaptation (RMDA) tested at 5° superior retina and contrast sensitivity (CS) emerged as being most strongly associated with AMD incidence and progression. ,,,
Given these findings, we hypothesize an association between rpeHRF and visual function, especially those associated with AMD incidence and progression, which may underscore rpeHRF’s potential value as a prognostic biomarker that may be used for clinical trial enrichment. We test this hypothesis leveraging the MACUSTAR data set.
As previously described in detail, , MACUSTAR is a multicenter, prospective, and low-interventional clinical study on subjects with AMD conducted across 20 sites in 7 European countries. The MACUSTAR study includes a battery of high-resolution retinal imaging modalities, visual function tests, and patient-reported outcome measures to identify biomarkers associated with AMD progression and to define novel end points for interventional trials. ,
METHODS
THE MACUSTAR STUDY AND PATIENT SELECTION
MACUSTAR (ClinicalTrials.gov Identifier: NCT03349801) adheres to the tenets of the Declaration of Helsinki, ethics approval was obtained from the local ethics committees at all participating study sites, and subjects provided written informed consent before study inclusion. ,,,,,,, These committees included University Hospital Bonn ethics committee (384/17), Paris Ouest IV (04/18_2), AIBILI (032/2017/AIBILI/CE), Nova Medical School (13507/2017), London Queen Square Research Ethics Committee (18/LO/0145), Center for Sundhed Glostrup (H-18000126), Comitato Etico Milano (37910/2018), Ospedale San Raffaele (dated 25/10/2018), Radboudumc technology center (2017–3954), and LUMC commissie medische ethiek (L18.055/SH/sh).
Patients were enrolled between March 2018 and February 2020, initially followed up every 6 months for 4 years, with ongoing yearly follow-up visits. Detailed inclusion and exclusion criteria have been previously described. Briefly, individuals between 55 and 85 years of age at baseline and of either normal macular health, eAMD, iAMD, or late AMD, as defined by the Beckman classification system, were eligible. , Exclusion criteria encompassed optic media opacities, preventing the acquisition of high-resolution retinal imaging, any confounding ocular conditions, such as diabetic retinopathy, uncontrolled glaucoma, participation in interventional trials, and prior interventions targeting AMD. For this specific analysis, only baseline visit data of normal-aged, eAMD, and iAMD eyes were included.
FUNCTIONAL TESTING
Patients underwent a battery of visual function assessments, including best corrected visual acuity (BCVA), low-luminance visual acuity (LLVA), Pelli-Robson CS, mesopic microperimetry (ie, pointwise sensitivity deviation; mesPSD), scotopic microperimetry (scoPSD), and RMDA testing.
BCVA and LLVA were assessed using standardized Early Treatment Diabetic Retinopathy Study (ETDRS) charts. For LLVA testing, a mesopic filter (reducing luminance to 3 cd/m 2) was put in front of the letter chart. CS was tested using Pelli-Robson charts at a 1-m distance.
Mesopic microperimetry was conducted using the S-MAIA (a modified macular integrity assessment microperimeter; iCare) after the study eye was dilated with tropicamide 0.5% and phenylephrine 2.5% and dark-adapted for 5 minutes. A stimulus grid of 33 points located at 2° intervals from 0° to 7° from fixation was used. An achromatic stimulus (Goldmann III) was presented for 200 ms using a 4-2 staircase strategy with a background luminance of 1.27 cd/m 2. After another 30 minutes of dark adaptation, scotopic microperimetry was performed using a red (627 nm) stimulus (Goldmann III) using the same strategy with no background illumination. Mesopic S-MAIA average (mean) threshold (mesPSD [dB]) and scotopic S-MAIA average (mean) threshold (scoPSD [dB]) were used for further analyses.
RMDA was assessed using the AdaptDX (OpZira) shortly after scotopic microperimetry. The study eye was bleached using a 0.25-ms flash at 8 × 10 4 cd/m 2 s, at 12° inferior retina along the vertical meridian (location of the test target). The stimulus was a 2°-diameter, 500-nm-wavelength circular target presented for 200 ms, beginning 15 seconds after the bleach. Participants were instructed to press a button once they saw a flashing light. Log thresholds were estimated using a modified, 3-1 staircase procedure. This procedure continued until either the rod-intercept time (RIT, the time taken for retinal sensitivity to recover to reach a threshold located within the second component of rod recovery [5 × 10 −3 scot cd/m 2, ie, 3 log units of stimulus attenuation]) was obtained or the test protocol ended (30 minutes). ,
RETINAL IMAGING
Following pupil dilatation, patients underwent a standardized retinal imaging protocol according to standard operating procedures by certified study site personnel at the baseline and every subsequent 6-monthly follow-up visit. Details of the imaging protocol have previously been described. , We limit the description of the imaging modalities to the ones relevant to the present analysis.
In brief, high-speed combined confocal scanning laser ophthalmoscopy and spectral domain (SD)-OCT using the Spectralis HRA + OCT device (Heidelberg Engineering; digital imaging resolution 768 × 768 pixels) comprised infrared reflectance (IR, 30° × 30°, automatic real-time mode [ART] mode ≥ 30 frames), fundus autofluorescence (30° × 30°, ART mode ≥ 30 frames), and SD-OCT imaging (30° × 25° enhanced depth-imaging mode, centered on the fovea [defined by the point of fixation], 241 B-scans, distance between scans 30 µm, automated real-time averaging mode 9 frames). Additionally, CFPs were acquired. OCT angiography (OCTA) imaging included 6 × 6-mm scans (Zeiss Cirrus HD-OCT 5000 Angioplex, Zeiss PLEX Elite 9000 swept-source OCT) or 20° × 20° raster scans (512 B-scans with 512 A-scans, centered on the fovea, ART 7 mode) performed with the Heidelberg Engineering Spectralis OCT-2 device.
SUBRETINAL DRUSENOID DEPOSIT, LH yper TD, RPEDC VOLUME, AND i HRF AREA ASSESSMENT
The detailed multimodal retinal imaging–based grading approach has been reported previously. All assessments were conducted at a centralized reading center (GRADE Reading Center Bonn, Bonn, Germany). The dense SD-OCT volume raster scan (241 B-scans) was considered, whereas additional retinal imaging modalities were referred to depending on the structural parameter investigated. For this study, the presence of subretinal drusenoid deposit (SDD) was evaluated as previously described. , The presence of LHyperTD (>250 µm diameter) was determined manually by examining the choroidal OCTA slab as previously described. ,
In addition, retinal pigment epithelium drusen complex (RPEDC) volume and iHRF area were quantitatively assessed using a deep learning–based approach. RPEDC volumes were collated from abnormal RPEDC thicknesses (≥3 SDs higher than the mean RPEDC thickness), as previously described in detail. , In brief, abnormal RPEDC volumes were defined as those exhibiting thickness values ≥3 SDs above the mean RPEDC thickness of the healthy control cohort, following prior established methodology. , iHRF were automatically segmented on each OCT B-scan using a deep learning algorithm trained on the MACUSTAR data set and subsequently internally validated on an independent data set. Python was employed as the programming language for these analyses, using various Python libraries to facilitate computations and support image generation, including OpenCV, NumPy, Polars, ImageIO, and Matplotlib.
rpe HRF GRADING
rpeHRF were graded using custom open-access FiJi software (available at ( https://sites.imagej.net/CreativeComputation/ ). After exporting the.XML files for each OCT volume from the Heidelberg Eye Explorer (HEYEX) software, the foveal center was defined in each volume as the highest peak of the external limiting membrane, where cone photoreceptors are at their longest, or as the lowest dip of the foveal valley (plug-in: “Find Fovea OCT”). , The “Mark HRF OCT” plug-in then simultaneously displays the IR image and the OCT volume. All eyes are shown as left eyes (ie, right eyes are flipped 180°) for consistency. Every B-scan was closely examined individually for the presence of rpeHRF.
Building on a recent publication by Beni and associates, rpeHRF were defined as hyperreflective lesions of similar reflectivity as the RPE where no clear distinction of the rpeHRF and the hyperreflectivity of the RPE band itself was possible, or when the lesion was considered to be a thickening of the RPE layer. In addition, choroidal hypotransmission needed to be visible posterior to the rpeHRF. Figure 1 shows examples of rpeHRF and iHRF. rpeHRF were then labeled on the B-scan, and the respective location on the IR image was simultaneously displayed by the plug-in. After the grading of 1 OCT volume was completed, the plug-in produced a.csv file that indicated each rpeHRF’s location in millimeters relative to the foveal center. All gradings were conducted by an experienced, reading center–trained grader (L.G.). A subset of 50 randomly selected eyes was assessed by a second experienced, reading center–trained grader (M.S.) to enable intergrader reliability statistics.
Examples of rpeHRF and iHRF. A1, B1. Infrared images of 2 independent iAMD patients. The green lines indicate the location of the OCT B-scan. A2, B2. Corresponding OCT B-scans. Red arrowheads highlight rpeHRF, and blue arrowheads highlight iHRF. iAMD = intermediate age-related macular degeneration, iHRF = intraretinal HRF, OCT = optical coherence tomography, rpeHRF = hyperreflective foci along the retinal pigment epithelium.
HEATMAP CREATION
All obtained.csv files were collated into one spreadsheet (Microsoft Excel; Microsoft Corp), where each rpeHRF is represented by a defined set of x and y coordinates (in mm) relative to the foveal center. We used MATLAB’s (version 9.5; The MathWorks) indwelling “heatmap” command to create a heatmap indicating the topographical rpeHRF distribution across all study eyes.
STATISTICAL ANALYSIS
All statistical analyses were performed using the R software environment (R Foundation, version 4.4.1). Continuous variables (rpeHRF count, HRF area, RPEDC volume, functional parameters) are presented as means and SDs. The prevalence of SDD is given as absolute numbers and relative frequencies. Cohen kappa ( k ) and intraclass correlation coefficients (ICCs) were used for intergrader repeatability statistics. Analysis of variance adjusted for age was used to compare all metrics between disease groups. Linear regression adjusting for age compared functional parameters between iAMD eyes with and without rpeHRF. Spearman r , adjusting for age, SDD presence, HRF area, and RPEDC volume, was used to evaluate the association between rpeHRF and visual function. All analyses were conducted on the respective complete cases.
RESULTS
COHORT CHARACTERISTICS
As summarized in Table 1 , 673 eyes from 673 patients (mean age 71.7 ± 6.9 years; 66.4% female) were included in this study. Out of these, 56 eyes were graded as normal aged (68.1 ± 6.4 years; 58.9% female), 34 as eAMD (71.7 ± 6.4 years; 79.4% female), and 583 as iAMD (72.0 ± 7.0 years; 66.4% female). In the normal group, rpeHRF were identified in 3 eyes (5.4%). Six eyes of the eAMD group (17.6%) and 280 eyes of the iAMD group (48.0%) showed rpeHRF. The mean number of rpeHRF across all eyes within each group was 0.16 ± 0.85 in normal aged eyes, 0.33 ± 0.96 in eAMD, and 1.61 ± 2.5 in iAMD eyes. Interreader reliability for rpeHRF assessment was almost perfect with 0.836 (κ) for rpeHRF presence and 0.823 (ICC; 95% CI 0.712-0.899, P <.001) for rpeHRF count. , LHyperTD were found in only 28 iAMD eyes (4.8%) and therefore not included in further analyses. As shown in Table 2 , after adjusting for age, a group difference was found between the AMD groups for rpeHRF, SDD, BCVA, LLVA, CS (all P <.001), and scoPSD ( P =.001), whereas no statistical difference was found for HRF area, RPEDC volume, RMDA, and mesPSD. Considering the descriptive overview of the functional measures ( Table 2 ), it can be concluded that rpeHRF and SDD were more frequent in iAMD compared with eAMD and normal aged eyes, whereas BCVA, LLVA, CS, and scoPSD were worse in iAMD compared with eAMD and normal aged eyes.
TABLE 1
Cohort Characteristics.
| Normal | eAMD | iAMD | All Eyes | |
|---|---|---|---|---|
| Number of eyes (%) | 56 (8.3) | 34 (5.1) | 583 (86.7) | 673 (100) |
| Age, y, mean (SD) | 68.1 (6.3) | 71.7 (6.4) | 72.0 (7.0) | 71.7 (6.9) |
| Female, n (%) | 33 (58.9) | 27 (79.4) | 387 (66.4) | 447 (66.4) |
| rpeHRF present, n (%) | 3 (5.4) | 6 (17.6) | 280 (48.0) | 289 (42.9) |
| No. of rpeHRF, mean (SD) | 0.16 (0.84) | 0.33 (0.96) | 1.61 (2.49) | 1.42 (2.39) |
| No. of iHRF, mean (SD) | — | 9.5 (10.6) | 33.6 (39.1) | 33.6 (39.0) |
eAMD = early age-related macular degeneration, iAMD = intermediate AMD, iHRF = intraretinal hyperreflective foci, rpeHRF = hyperreflective foci contiguous with the retinal pigment epithelium.
TABLE 2
Differences in Structural and Functional Parameters Between Study Groups.
| Parameter | No. of Eyes | Normal | eAMD | iAMD | P |
|---|---|---|---|---|---|
| Number of rpeHRF, mean (SD) | 647 | 0.16 (0.84) | 0.33 (0.96) | 1.61 (2.49) | <.001 |
| SDD present, n (%) | 671 | 0 (0.0) | 4 (11.8) | 157 (26.9) | <.001 |
| HRF area, mm 2, mean (SD) | 586 | NA | 0.0311 (0.0364) | 0.1114 (0.1852) | .43 |
| RPEDC volume, mm 3, mean (SD) | 580 | NA | 0.0007 (0.0007) | 0.0132 (0.0321) | .60 |
| BCVA, logMAR, mean (SD) | 672 | –0.04 (0.08) | 0.01 (0.08) | 0.03 (0.10) | <.001 |
| LLVA, logMAR, mean (SD) | 671 | 0.14 (0.09) | 0.19 (0.14) | 0.24 (0.16) | <.001 |
| RIT, min, mean (SD) | 480 | 4.24 (1.26) | 8.23 (15.83) | 7.65 (8.47) | .11 |
| CS, logCS, mean (SD) | 671 | 1.71 (0.16) | 1.63 (0.16) | 1.55 (0.178) | <.001 |
| mesPSD, db, mean (SD) | 614 | 2.20 (0.88) | 2.63 (1.18) | 2.79 (1.53) | .09 |
| scoPSD, db, mean (SD) | 583 | 2.31 (0.68) | 2.90 (1.33) | 3.12 (1.39) | .001 |
eAMD = early age-related macular degeneration, BCVA = best corrected visual acuity, CS = contrast sensitivity, iAMD = intermediate age-related macular degeneration, LLVA = low-luminance visual acuity, mesPSD = mesopic pattern standard deviation, RPEDC = retinal pigment epithelium drusen complex, rpeHRF = Hyperreflective foci along the retinal pigment epithelium, RIT = rod intercept time, scoPSD = scotopic pattern deviation, SDD = subretinal drusenoid deposit.
Bold letters indicate statistical significance.
STRUCTURAL AND FUNCTIONAL DIFFERENCES IN i AMD EYES DEPENDING ON rpe HRF PRESENCE
The subgroup analysis limited to iAMD eyes revealed significant differences in SDD presence, iHRF area, and RPEDC volume between iAMD eyes with and without rpeHRF (all P <.001). BCVA, LLVA, CS, scoPSD (all P <.001), and mesPSD ( P =.015) also differed significantly between eyes with and without rpeHRF. No difference was found for RMDA. Details of this analysis are shown in Table 3 .
TABLE 3
Differences in Structural and Functional Parameters Between iAMD Eyes With and Without rpeHRF.
| No. of Eyes | Estimate | SE | P | |
|---|---|---|---|---|
| SDD present | 556 | –1.000 | 0.206 | <.001 |
| HRF area | 558 | –0.1478 | 0.015 | <.001 |
| RPEDC volume | 552 | –0.016 | 0.003 | <.001 |
| BCVA | 557 | –0.041 | 0.008 | <.001 |
| LLVA | 556 | –0.065 | 0.012 | <.001 |
| RIT | 390 | –1.096 | 0.869 | .203 |
| CS | 556 | 0.078 | 0.014 | <.001 |
| mesPSD | 506 | –0.331 | 0.136 | .015 |
| scoPSD | 477 | –0.633 | 0.123 | <.001 |
iAMD = intermediate AMD, rpeHRF = Hyperreflective foci along the retinal pigment epithelium, SDD = subretinal drusenoid deposit,
RPEDC = retinal pigment epithelium drusen complex, BCVA = best corrected visual acuity, LLVA = low-luminance visual acuity, RIT = rod intercept time, CS = contrast sensitivity, mesPSD = mesopic pattern standard deviation, scoPSD = scotopic pattern deviation
Bold letters indicate statistical significance.
ASSOCIATION OF rpe HRF COUNT AND VISUAL FUNCTION
As shown in Table 4 , BCVA ( r = 0.10, P =.016), LLVA ( r = 0.11, P =.011), CS ( r =-0.19, P <.001), and scoPSD ( r = 0.13, P <.01) were significantly associated with rpeHRF count. There was no significant association for RMDA and mesPSD.
TABLE 4
Association of rpeHRF Count and Visual Function.
| Spearman r | P | |
|---|---|---|
| BCVA | 0.10 | .016 |
| LLVA | 0.11 | .011 |
| RIT | –0.02 | .75 |
| CS | –0.19 | <.001 |
| mesPSD | 0.08 | .07 |
| scoPSD | 0.13 | <.01 |
BCVA = best corrected visual acuity, CS = contrast sensitivity, LLVA = low-luminance visual acuity, mesPSD = mesopic pattern standard deviation, RIT = rod intercept time, rpeHRF = hyperreflective foci along the retinal pigment epithelium, scoPSD = scotopic pattern deviation.
Bold letters indicate statistical significance.
rpe HRF TOPOGRAPHY
Figure 2 shows a heatmap of the topographic distribution of all rpeHRF of all study eyes. The gray circles indicate the central subfield and the inner and outer ring of the Early Treatment Diabetic Retinopathy Study (ETDRS) grid. Colors leaning toward red indicate higher numbers of rpeHRF at the specific location, whereas dark blue colors indicate low numbers of rpeHRF. rpeHRF cluster within the central subfield of the ETDRS grid, less so in the inner ETDRS ring, and even less inside the outer ETDRS ring. Only few rpeHRF are located outside of the ETDRS grid.
Topographic rpeHRF distribution pattern. A heatmap of topographic rpeHRF distribution is presented. All eyes were graded as left eyes (ie, right eyes were flipped) for uniform presentation. The rings represent the central subfield and inner and outer rings of the Early Treatment Diabetic Retinopathy Study (ETDRS) grid. x and y axes show distances relative to the foveal center. Colors indicate the amount of rpeHRF at each location. rpeHRF are most abundant within the central subfield followed by the inner and outer ring. Only a few were identified outside the ETDRS grid. rpeHRF = hyperreflective foci along the retinal pigment epithelium.
DISCUSSION
In our analysis, rpeHRF were most abundant in iAMD, followed by eAMD eyes, and typically clustered around the foveal center. All structural biomarkers indicating a stronger disease load (SDD presence, iHRF area, RPEDC volume), and all (BCVA, LLVA, CS, mesPSD, scoPSD) but 1 (RMDA) of the visual function tests were worse in iAMD eyes in the presence of rpeHRF. rpeHRF count negatively impacted BCVA, LLVA, CS, and scoPSD, yet the associations were weak.
Our findings confirm our hypothesis of rpeHRF presence and count being associated with visual function impairment. These associations were weaker compared with recent publications from a different cohort, where correlation coefficients for fluorescence lifetime imaging ophthalmoscopy decay rates, for example, measured in the outer ETDRS ring reach up to 0.68 for RMDA (tested at 5°) and–0.25 for mesopic CS. ,,,, However, one needs to acknowledge that in their analysis, the authors used a different AMD classification system (AREDS 9-step ) and did not adjust for other imaging biomarkers such as RPEDC volume, SDD presence, and iHRF area, as we did herein. Other groups that consider the presence of potentially confounding OCT biomarkers generally report associations of similar strengths to our results. ,,
Previous research reliably identified the strongest association between structural biomarkers and visual function for RMDA. ,,,,, Interestingly, in our sample, rpeHRF count was not associated with RMDA. Possible explanations may be that as a result of the multicenter study design of MACUSTAR, fewer RMDA data points were available compared with other investigated visual functions. In addition, the strong associations of RMDA and retinal structure were identified when RMDA was tested at 5° superior retina. The test target location at 12° eccentricity in the MACUSTAR study corresponds to the region of peak rod density, rendering it an appropriate choice. On the other hand, parafoveal (eg, at 5°) rod photoreceptors were identified to decline in number more strongly in aging and AMD compared with their counterparts at greater eccentricities, which may amplify age- and disease-related functional impairments. ,
In general, rod recovery was found to be slower near the foveal center compared to more eccentric testing locations. ,,,, Given the significant association between rpeHRF count and scoPSD, we believe that rpeHRF are still a reliable indicator of rod dysfunction. scoPSD was shown to be impaired in the presence of SDD and large sub-RPE drusen, for which we adjust in our analysis approach, which emphasizes the independent association we found for rpeHRF count. ,
CS has been linked to AMD onset and progression, , and associated with RPEDC volume, iHRF area, and SDD. Despite the significant association between rpeHRF and CS being weak in our sample, we suggest that it is still clinically meaningful, given that we adjust for the aforementioned biomarkers, which reduces the strength of the identified association. We show that visual function in iAMD eyes in which rpeHRF are present is worse compared to iAMD eyes without rpeHRF, which indicates that these eyes are more advanced within the iAMD spectrum. These findings, in combination with Berni and associates’ results, that rpeHRF independently predict LHyperTD onset, underscore rpeHRF potential value as a prognostic biomarker and enrichment factor for clinical trials.
In our sample, rpeHRF were more abundant in iAMD, compared with eAMD and healthy aged eyes, which aligns with other OCT biomarkers being more prevalent in more advanced disease. ,,,, Although this finding is intuitive, it should be examined more closely. As mentioned above, rpeHRF, along with iHRF, typically correspond to focal hyperpigmentation on CFP. ,, The MACUSTAR study uses the Beckman classification to assign AMD stages, which includes all eyes showing pigmentary abnormalities in the iAMD group. Previous work comparing the AREDS 9-step severity scale to the Beckmann classification system showed that approximately 40% of AREDS 9-step eAMD eyes were considered iAMD in Beckman.
It is worthwhile to consider the different origins of both classification systems: the AREDS 9-step system relies on CFP data from 3212 eyes, using a classification and regression tree analysis to rate various CFP features’ impact on disease severity. , The Beckman classification system is derived from Delphi process by clinical experts who discussed clinically, that is, funduscopically, visible disease features associated with disease progression. The uneven rpeHRF distribution between eAMD and iAMD eyes may hence be partially attributable to the used staging system.
As a result of the yet-to-be completely understood multifactorial pathophysiology of AMD, RPE cells are thought to experience progressive dysfunction and eventually epithelial-mesenchymal transition. ,,, This modification is likely at the origin of the RPE cells’ migration toward the inner retina, where they can be identified as iHRF. Epithelial-mesenchymal transition may start before the cells leave the monolayer, which indicates that rpeHRF formation may be a very early biomarker for RPE pathology.
Our study shows that rpeHRF concentrate at the foveal center and are more sparsely distributed at higher eccentricities, which mirrors the distribution of cone photoreceptors in the human retina. This distribution pattern is shared by other AMD-related findings, such as sub-RPE drusen, ,, choriocapillaris flow deficits, and LHyperTD. In addition, macular xanthophyll carotenoids (lutein, zeaxanthin, and mesozeaxanthin) also accumulate within the central ETDRS subfield and have been linked to AMD pathology. ,
The rpeHRF distribution can be further compared to the distribution of iHRF. Although rpeHRF are most abundant inside the central ETDRS subfield, iHRF are more commonly found inside the inner ring of the ETDRS grid. Provided that rpeHRF are precursor lesions of iHRF, the RPE cells’ migration pattern appears centrifugal. Hypoxia has been identified as a key driver of RPE epithelial-mesenchymal transition. , The foveal avascular zone has a similar area as the central ETDRS grid subfield. ,, We therefore suggest that the different topography of rpeHRF and iHRF reflects a distinct migration pattern toward intraretinal vessels. Ongoing longitudinal analyses of the same data set will explore this hypothesis further.
Several limitations need to be acknowledged when interpreting our data. We used SD-OCT B-scans to manually grade rpeHRF, which differs from the semiautomated approach Berni and associates employed using swept-source OCT scans. In addition, Berni and associates used rpeHRF area for statistical analyses, whereas we rely on rpeHRF count. However, our method has been established and verified by assessing iHRF in a different sample, which legitimizes our approach. We herein exclusively use cross-sectional data, which does not allow to address the prognostic value of rpeHRF. The MACUSTAR data set mostly comprises iAMD eyes, with much fewer normal or eAMD eyes, which hinders the interpretability of our data regarding these groups. Currently, associations between rpeHRF and pigmentary changes on CFP are not included in our analysis. Major strengths of our study include the large, longitudinally acquired and standardized data set, and the robust statistical approach. Disease grading and biomarker stratification are conducted in a central reading center setting, following strict predefined standard operating procedures.
In conclusion, this study shows that rpeHRF are independently associated with visual function impairment indicative of AMD progression, underscoring their potential value as a prognostic biomarker and enrichment factor for clinical trials. rpeHRF topography fits well into the framework of AMD pathology, supporting RPE epithelial-mesenchymal transition as one of its features. Ongoing longitudinal analyses focus on rpeHRF’s prognostic value for iAMD progression to late-stage disease manifestations and will further clarify rpeHRF’s presumed migration pattern toward intraretinal vessels located in the inner ring of the ETDRS grid.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
