Aims To describe the vascular changes in patients affected by non-neovascular age-related macular degeneration (AMD), featuring reticular pseudodrusen (RPD), drusen, or both RPD and drusen by means of optical coherence tomography angiography (OCT-A).
Methods Cross-sectional observational case series. Patients with non-neovascular AMD presenting at the Medical Retina Service of the Department of Ophthalmology, University Vita-Salute San Raffaele in Milan were recruited. Patients underwent best-corrected visual acuity, biomicroscopy, infrared reflectance, short-wavelength fundus autofluorescence and OCT-A (AngioPlex, CIRRUS HD-OCT 5000, Carl Zeiss Meditech, Dublin, USA). Main outcome was quantification of vessel density, stromal tissue, and vascular/stromal (V/S) ratio at the choriocapillaris (CC), the Sattler and Haller's and the whole choroid layers among different groups of patients with non-neovascular AMD by means of binarised OCT-A scans.
Results 45 eyes of 34 patients were enrolled (15 eyes of 11 patients with RPD, group 1; 15 eyes of 11 patients with drusen, group 2; 15 eyes of 12 patients with mixed phenotype, group 3). The CC, the Sattler and Haller's and the whole choroid vessel density were reduced in all groups of patients (p=0.023, p=0.007 and p=0.011 in group 1, group 2 and group 3 for the CC; p=0.021, p=0.037 and p=0.043 in group 1, group 2 and group 3 for the Sattler and Haller's density; p=0.016, p=0.002 and p<0.001 in group 1, group 2 and group 3 for the choroidal density), with significantly lower V/S ratios compared with healthy controls.
Conclusions Patients with non-neovascular AMD show significant choroidal vascular depletion and fibrotic replacement, suggesting a possible role in the pathogenesis and progression of the disease.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Age-related macular degeneration (AMD) is the major cause of visual impairment in adults older than 50 years in Western society.1 The Age-Related Eye Disease Study Group has classified AMD into different categories (non-neovascular and neovascular) on the basis of the presence of choroidal neovascularisation (CNV).2 ,3
Non-neovascular AMD has been defined by the presence of retinal pigment epithelium (RPE) alterations and Bruch's membrane (BM) degenerative changes, with the deposition of focal deposits of lipofuscin, photoreceptor debris and inflammatory components between the BM and the RPE (drusen).3 ,4 Recently, reticular pseudodrusen (RPD) has been identified as an additional phenotype, strongly associated with AMD.5 ,6 The term ‘pseudodrusen visible en lumière bleue’ (‘pseudodrusen visible in blue light’) was introduced by Mimoun and associates5 in 1990 to define ‘retinal lesions with a variable diameter of about 100 microns that did not appear hyperfluorescent on fluorescein angiography’. Both drusen and RPD have been recognised as independent clinical risk factors for progression to a more advanced stage of AMD.6 Furthermore, these lesions are known to affect photoreceptor function, in particular, rods.7
AMD development and progression depend mostly on the complex interaction between the genetic, metabolic and environmental factors that affect the different structures of the macular region. Histopathology studies have demonstrated that most of these changes involve the outer retina and the choroidal cytoarchitecture, where extracellular matrix is replaced by fibrosis, leucocyte infiltration and oedema.8 ,9
Fluorescein angiography,10 indocyanine green angiography,11 fundus autofluorescence (FAF)12 and enhanced depth imaging optical coherence tomography (EDI OCT)13 have been all used to quantify stromal changes and vascular alterations in patients with AMD.
Optical coherence tomography angiography (OCT-A) is a relatively new, dyeless, depth-resolved technique that allows the visualisation of retinal and choroidal microvasculature by detecting erythrocyte flow within the retinal capillaries.14 ,15 Two major motion contrast techniques, phase-based and amplitude-based, are used to render retinal and choroidal depth imaging combined with ‘en face’ OCT-derived imaging. Digitally binarised OCT-A scans through open access software, that is, ImageJ, may be used to non-invasively calculate the extent of the vascular network and, consequently, of the extracellular tissue in comparison with the total analysed area.16 ,17 The purpose of this study is to quantify the vascular changes at the choriocapillaris (CC) and choroidal layers in patients affected by drusen and RPD, by means of OCT-A.
Material and methods
This is a clinical practice, observational, cross-sectional study. A consecutive series of patients with non-neovascular AMD who presented to the Medical Retina & Imaging Unit of San Raffaele Hospital in Milan between October 2015 and June 2016 were enrolled. All patients signed a written general consent to participate to observational studies, which was approved by the ethics committee of San Raffaele Hospital, and the procedures were performed according to the tenets of the Declaration of Helsinki.
Inclusion criteria were age ≥55 years and diagnosis of non-neovascular AMD, along with clear dioptric to allow high-quality OCT-A examination. All the patients affected with any other ocular disorder, along with those with any signs of RPE atrophy, advanced form of AMD (including geographical atrophy or CNV), or those previously undergoing intravitreal, photodynamic or laser treatment in the study eye, were excluded from the analysis. Subjects with myopic refractive error >−6 D and axial length >26 mm were also excluded.18 Uncontrolled systemic blood pressure, diabetes mellitus and peripheral vasculopathy were also considered as exclusion criteria. Fifteen healthy age-matched patients (15 eyes included in the analysis, 1 eye for each patient) without any ocular or systemic disease acted as a control group.
Each patient underwent best-corrected visual acuity (BCVA), biomicroscopy, applanation tonometry, multicolour imaging, infrared (IR) reflectance, short-wavelength (488 nm) fundus autofluorescence (SW-FAF) (Spectralis, Heidelberg Retinal Angiography, Heidelberg, Heidelberg, Germany), spectral domain optical coherence tomography (SD-OCT) and OCT-A. On the basis of clinical examination, patients were divided into three groups: (1) those with RPD (confirmed on ophthalmoscopy, multicolour, IR reflectance and SW-FAF) not accompanied by drusen; (2) those with small (<63 μm) or medium–large drusen (63–124 μm) within the macula, without RPD and (3) those presenting both RPD and drusen, categorised as ‘mixed’.
Horizontal and vertical structural SD-OCT lines centred on the fovea were obtained for each patient (each with 50 averaged OCT-B scans −1024 A scans per line) using the EDI technique, as reported previously.19 The choroidal thickness (CT), defined as the vertical distance between the hyper-reflective line of BM and the chorioscleral interface, was manually measured under the foveal depression, and at 500, 1000 and 1500 μm intervals nasally, temporally, superiorly and inferiorly to the fovea, and then averaged by two independent trained retina specialists (MVC and AR). Examinations were performed after pupil dilatation in the same time frame for all subjects, between 9:00 and 13:00 hours, to avoid bias because of diurnal variations in CT.20 The CT was compared among groups.
Measurement of outer retinal layers’ thickness was obtained by the automated segmentation protocol of the Spectralis OCT (Heidelberg Eye Explorer V.18.104.22.168) to acquire extra information about patients' retinal morphology. The layers included in the analysis were: (1) RPE–photoreceptors (outer segments of photoreceptors, ellipsoid zone and myoid zone) determined from the hyper-reflective BM to the hyper-reflective external limiting membrane (ELM) and indicated as ‘outer retinal layers’ by the software; (2) outer nuclear layer (ONL) from the ELM to the outer border of the hyper-reflective band of the outer plexiform layer (OPL); (3) OPL from the inner border of the ONL to the outer border of the hyporeflective band corresponding to the inner nuclear layer (INL).21 Layer segmentation errors were manually adjusted before image processing. Thickness measurements for these retinal layers were performed in the nine sectors of the ETDRS macular grid at 1, 3 and 6 mm.
OCT-A was performed by means of Zeiss AngioPlex, CIRRUS HD-OCT models 5000 (Carl Zeiss Meditech, Dublin, USA), relying on optical microangiography algorithm; 3×3 mm OCT angiograms were recorded for each patient. Only images with signal quality >6 were included in the analysis. The segmentation of the CC, of the Sattler and Haller's layers and of the whole choroid (CC+Sattler and Haller's layers) was manually defined as the vertical distance between the hyper-reflective line of the RPE (by selecting the ‘RPE-fit’ option on AngioPlex) and the BM (by selecting the ‘RPE’ option on AngioPlex), between the BM and the chorioscleral interface and between the RPE and the chorioscleral interface, respectively. Two trained retina specialists (MVC and AR) independently performed the qualitative analysis of the images for segmentation.
All 3×3 OCT-A images were exported from the system as a Joint Photographic Experts Groupfile and were analysed through the National Institutes of Health ImageJ 1.50 (National Institutes of Health, Bethesda, Maryland, USA) software. Each image was converted from 8-bit image into red green blue colour type, and then was split into the three channels (red, green and blue); the red channel was chosen as the reference. The adjust→automated threshold plugin set to ‘default’ was applied (a variation of the ISODATA algorithm); the ‘white objects on black background’ option was kept selected.16 This plugin binarises 8-bit and 16-bit images using various global (histogram-derived) thresholding methods. The segmented phase is always shown as white (255). The white pixels were defined as the luminal area, and the dark pixels were defined as the interstitial choroid or choroidal stroma. Vessel density was expressed as the proportion between the white pixels and the total area; stromal density, expressed as 1−vessel density, was introduced in order to calculate the vascular/stromal ratio (V/S ratio). The CC, the Sattler and Haller's layers and the whole choroid (CC+Sattler and Haller's layers) were analysed using this method; the variables of each patients' group were compared with the other groups and controls for the three different layers of segmentation.
Statistical analysis, including descriptive statistics for demographics and main clinical records, and comparative analysis (Student's t-test analysis for independent samples and one-way analysis of variance with Bonferroni correction) were performed through GraphPad Prism V.5.0 (GraphPad software, San Diego, California, USA). BCVA was converted to the logarithm of the minimum angle of resolution (LogMAR) for statistical purpose. The inter-rater agreement was assessed using a two-way mixed consistency intraclass correlation (ICC) to assess the degree of consistency between readers (MVC and AR) in their ratings of the CT and of the OCT-A segmentation across subjects. The chosen level of statistical significance was two-sided p<0.05.
A total of 45 eyes of 34 patients (16 males, 47.1%) were enrolled in the study (15 eyes of 11 patients with RPD, group 1; 15 eyes of 11 patients with drusen, group 2; 15 eyes of 12 patients with mixed phenotype, group 3) (figures 1 ⇓–3). Mean age was 74.2±5.9 (62–83). Fifteen eyes of fifteen healthy age-matched and sex-matched subjects (7 males, 46.7%), mean age 72.6±7.6, were included as controls (patients vs controls, p>0.9). Mean BCVA was 0.12±0.08 LogMAR in the study population and 0.03±0.01 LogMAR in the control group; mean refractive error was 0±2.0 D spherical equivalent. No meaningful statistical difference was found in BCVA among the study groups (p>0.9). Demographics and main clinical characteristics of the patients are listed in table 1.
Global CT within the macular area (3×3 mm) turned out to be severely reduced in patients with RPD with respect to other categories and controls (p<0.001) (table 2).
Outer retinal layer thickness (RPE–photoreceptors, ONL and OPL) has been analysed: the RPE–photoreceptors layer showed statistically significant difference for patients with drusen (with or without RPD) compared with controls (p=0.006), whereas the ONL and OPL did not show any significant difference among groups (see online supplementary table S1).
Outer retinal layer thickness of study population evaluated on structural spectral domain optical coherence tomography (OCT).
The CC, the Sattler and Haller's and the choroidal vessel density, calculated after binarisation of OCT-A images, were reduced in all groups of patients compared with healthy controls (p=0.023, p=0.007 and p=0.011 in group 1, group 2 and group 3 for the CC; p=0.021, p=0.037 and p=0.043 in group 1, group 2 and group 3 for the Sattler and Haller's density; p=0.016, p=0.002 and p<0.001 in group 1, group 2 and group 3 for the choroidal density) (table 2). The stromal density (1−vessel density) disclosed a similar behaviour (table 2). Pooling together the patients featuring RPD (group 1 and group 3) increased the statistical meaningfulness of the difference found between patients and controls. However, no differences were found when compared with patients featuring drusen only (group 2).
The V/S ratio at the CC, at the Sattler and Haller's layer and at the entire choroidal layer was significantly reduced in all patient groups (p=0.017, p=0.006 and p=0.008 in the group 1, group 2 and group 3, respectively for the CC; p=0.026, p=0.026 and p=0.045 in the group 1, group 2 and group 3, respectively for the Sattler and Haller's layer; p=0.005, p<0.001 and p<0.001 in the group 1, group 2 and group 3, respectively for the choroidal layer) (figure 4 and table 2).
ICC was in the excellent range (ICC=0.94 for CT and ICC=0.96 for OCT-A segmentation), indicating that readers had a high degree of agreement and suggesting that variables were rated similarly across readers. CT measurements and OCT-A segmentation were therefore deemed to be suitable for use in the hypothesis tests of the present study.
This study was designed to determine by means of a valid, objective and reproducible method, the choroidal luminal and the interstitial area on OCT-A images in patients presenting different features of non-neovascular AMD.
The use of advanced imaging technologies has led to a deeper understanding of the choroidal physiological characteristics and its pathological changes in several ocular conditions, especially in those where ocular perfusion is primarily impaired, such as chronic open-angle glaucoma,22 diabetes,23 uveitis24 and AMD.25 Healthy choroid receives about 70% of total eye blood flow, featuring the highest blood flow/unit weight in the human body. Structurally, it consists of a complex network of blood vessels and extracellular matrix, containing immune system cells, pigment cells, smooth muscle, collagen bundles and intrinsic neurons forming the choroidal parasympathetic plexus, that accomplish fundamental physiological functions: providing nourishment to the outer layers of the retina and of the optic nerve head, acting as heat regulator, absorbing photons to reduce light scattering and aiding in bulb length adjustment during accommodation.26 ,27
Previous studies have demonstrated that global choroidal thickness tends to decrease with ageing, and this might play a role in the pathophysiology of AMD. Moreover, a significant decrease in thickness and perfusion in the human CC has been correlated with the extent of drusen and RPD in patients suffering from non-neovascular AMD.9 ,28 Accordingly, with these findings, we have found that eyes with RPD show a significative reduction of the mean total choroidal thickness on SD-OCT, especially in the extrafoveal quadrants of the macula.
Digital binarisation represents an objective and reproducible method to quantify the changes taking place at the choroidal level. Binarisation has been validated to convert a ‘noised’ image in white/black dichotomy, applying an automatic threshold to the original greyscale.29 Through this method, we calculated the vessel extent in the CC, in the Sattler and Haller's layers and in the whole choroid in different groups of patients with non-neovascular AMD. The vessel density turned out to be significantly lower at the three different layers of segmentation, in the RPD, in the drusen and the mixed phenotypes, with a particular involvement of the CC. Moreover, the proportion between vessels and stroma turned out to be significantly lower in all patient groups with respect to controls, with a relative reduction of the vascular density in favour of the stromal components.
These findings confirm that important choroidal vascular depletion takes place in patients with RPD and drusen: in the early stages of AMD, the choroidal layer could be still normal in thickness, but already shows dramatic alterations in its composition, with a predominance of the stromal tissue on the vascular network. This tendency appeared to be numerically more pronounced at the CC rather than the deeper choroidal layers (Sattler and Haller's), being meaningful at both levels.
Current OCT-A results are in line with previously published literature. In fact, in patients with non-neovascular AMD, histopathological studies and multimodal imaging through swept-source OCT and EDI-OCT have shown a general thinning of the choroid along with small vessel loss and enlarged avascular spacing between the choroidal vessels, due to increased stroma deposition30–33 Recently, digital binarisation of structural EDI SD-OCT has been used to confirm these data, calculating the choroidal luminal and the stromal area in eyes with drusen and RPD. Corvi and associates have found an overall thinning of the choroidal thickness in eyes with drusen and RPD, with a progressive loss of vascularity and its consecutive stromal replacement in a 24-month follow-up.13
Choroidal vascular depletion could have significant clinical implications: the outer retinal layers do not have vascular supply, and the choroid is the major source of oxygen and nutrients to the RPE–photoreceptor layers. In conditions where the choroid is thinner, such as pathological myopia, a positive correlation between macular CT and outer retinal layers has been found.18 More recently, Abdolrahimzadeh and associates have demonstrated a positive relationship between inner retinal layers (inner nuclear layer, ganglion cell complex and inner plexiform layer) and CT, even if the inner retinal layers do not depend on vascular support from the choroidal system and have a separate vascular network.34 Therefore, reduced choroidal supply could influence all the retinal thickness and lead to altered trophism of all retinal layers. In our analysis, outer retinal layer thickness (ONL and OPL) did not show any significant difference among groups. The RPE–photoreceptors layer showed significant difference for patients with drusen (with or without RPD) compared with controls; however, we cannot exclude that such difference resulted from incorrect measurements due to drusen material under the RPE that increases its overall thickness.
OCT-A is a nascent technology that offers the chance to non-invasively investigate the CC and the choroid separately. Only few studies have focused on the composition of CC and choroid in patients featuring early stages of AMD, especially RPD.35 ,36 Alten and associates have found a more severe reduction in vessel density measured on 10 μm slab OCT-A (including only CC) than on 30 μm slab OCT-A (including both the CC and the choroidal layers). They reported that in eyes affected with RPD, functional changes in vessel architecture take place at the CC layer in particular, rather than in the larger choroidal vessels. Similarly, we found a reduction that is numerically more significant in the choroidal thickness, in the vessel density and in the V/S ratio at the CC, compared with the other classes of non-vascular AMD. This underscores the important role of the CC in RPD pathophysiology.33 Further analyses are required to correlate OCT-A to histological findings in normal and pathological eyes.
Limitations of this study include the small sample size and the lack of follow-up. Small sample is mainly related to the strict inclusion criteria for drusen, RPD and control groups, and the exclusion of patients with incipient atrophy, which would have created significant artefacts on OCT-A quantification. Another potential limitation is that we have assumed the variables’ distribution of the analysis as normal, without a generalised linear model to adjust for intraeye parameters within the same patient. Strength of the study is the high reproducibility of the study, supporting the statistical meaningfulness of the results.
In conclusion, we converted OCT-A scans to binary images to quantify the luminal and interstitial space ratios of the choroid in different categories of patients with AMD. Using this technique, we found increased vessel depletion in patients with non-neovascular AMD compared with normal control subjects. This method might provide deeper knowledge in the pathophysiology of human choroid in non-neovascular AMD, and possibly allows predicting the natural history of the disease in an individualised therapeutic approach. In particular, non-invasive in vivo identification of the histopathological substrate and understanding of biogenesis of non-neovascular AMD will open the door to further therapeutic targets in its treatment.
Contributors All the authors contributed to the conception or design of the work, the acquisition, analysis and interpretation of data, drafting the work, revising it critically for important intellectual content and gave final approval of the version to be published.
Competing interests FB consultant for Alcon (Fort Worth,Texas,USA), Alimera Sciences (Alpharetta, Georgia, USA), Allergan Inc (Irvine, California,USA), Farmila-Thea (Clermont-Ferrand, France), Bausch And Lomb (Rochester, New York, USA), Genentech (San Francisco, California, USA), Hoffmann-La-Roche (Basel, Switzerland), Novagali Pharma (Évry, France), Novartis (Basel, Switzerland), Bayer Shering-Pharma (Berlin, Germany), Sanofi-Aventis (Paris, France), Thrombogenics (Heverlee,Belgium), Zeiss (Dublin, USA), Pfizer (New York, USA), Santen (Osaka, Japan), Sifi (Aci Sant'Antonio, Italy). GQ consultant for: Alcon (Fort Worth, Texas, USA), Alimera Sciences (Alpharetta, Georgia, USA), Allergan (Irvine, California,USA), Bayer Shering-Pharma (Berlin, Germany), Bausch and Lomb (Rochester, New York, USA), Molecular partners (Zurich, Switzerland), Novartis (Basel, Switzerland), Ophthotech (New York, New York, USA)
Patient consent Obtained.
Ethics approval Ethics committee of San Raffaele Hospital.
Provenance and peer review Not commissioned; externally peer reviewed.