Using high-speed high resolution SDOCT imaging, retinal layers can now be measured in living eyes with non-neovascular AMD, with a precision adequate to measure focal abnormalities over drusen. In the series of AMD eyes studied here, there was a significant and profound decrease in thickness of the PRL focally over drusen compared to age-matched control eyes, while the inner retinal layers were unaffected. The decrease in PRL correlated well with druse height and minimally with druse width. Although the difference in PRL area between AMD and control eyes was not significant (perhaps due to the very small area occupied by drusen relative to the total PRL area), the increase in RPE area (which included drusen) correlated with a decrease in PRL area in the AMD eyes. Despite the PRL changes over drusen, the mean area of the PRL layer did not differ between AMD and control eyes. This analysis supports an association between drusen and focal loss of the photoreceptor layer in eyes with AMD but suggests a lack of widespread photoreceptor loss (although there could be loss at a level not detectable in this small sample).
This finding of in vivo
PRL thinning over drusen with preservation of the inner retinal layers correlates with previously published postmortem studies of photoreceptor loss in non-exudative AMD. Curcio et al identified moderate photoreceptor loss in the parafovea, with rod loss greater than cone loss at the same location.8, 27
Focal loss of photoreceptors was not recorded over specific drusen; however, the photoreceptor loss did appear more widespread than noted in our in vivo
study. The postmortem finding of no difference in the ganglion cell layer between the non-exudative AMD eyes and controls agrees with our finding of comparable inner retinal layer thickness between AMD and control eyes. Our findings of decrease in PRL thickness by approximately 25% over drusen also compares favorably to the Johnson et al.9
postmortem observation that photoreceptor density was consistently decreased (mean of approximately 30%) over drusen. The authors divided drusen into four categories ranging from extremely small (31–50 μm) to the largest of 91 μm diameter and over. Although photoreceptor cell density progressively decreased as druse width increased in the postmortem study of smaller drusen, in this in vivo
SDOCT study of eyes with drusen greater than 125 μm, the photoreceptor layer thickness correlated closely to druse height and not as closely with druse width.
Note that the width of some of the drusen in this study clearly exceeded the typical maximum size on color fundus photograph of up to 434 μm (based on the diameter of the I-2 ring and an optic nerve diameter of 1800 μm) after which it is more likely to be classified as drusenoid pigment epithelial detachment (there is no specific size endpoint between a druse, coalescent drusen, and drusenoid pigment epithelial detachment). In the AREDS grading system of AMD phenotypes,17
drusen are categorized by a hard or soft appearance based on “uniformity of density, sharpness of edges and thickness.”17
and maximum diameter is also recorded. Drusenoid pigment epithelial detachment is differentiated from drusen and scored as questionable, present or absent, but size is not recorded. The differentiation between confluent drusen (margins in contact) and drusenoid PED is not explicitly defined in phenotyping studies, but becomes more obvious in cross-sectional imaging with SDOCT with lateral measurement of drusen. In the SDOCT images, there are multiple occurrences of drusen with confluent margins, resulting in a large lateral dimension for the drusen although there is no single broad domed lesion (). Toth et al. reported a greater lateral extent of drusen identified on SDOCT compared to corresponding color photographs for level 3 AMD eyes (Toth C.A., Farsiu S., Chiu S.J., Khanifar A.A., Izatt J.A., Automatic Drusen Segmentation and Characterization in Spectral Domain Optical coherence Tomography (SDOCT) Images of AMD Eyes. Paper presented at: ARVO annual meeting, May 1, 2008; Fort Lauderdale, FL). In a separate publication, we are addressing a system of categorizing drusen by size, shape and composition so as to relate the SDOCT findings to conventional color fundus images.23
In addition to the quantitative changes in the retina over drusen, with high-speed high-resolution SDOCT imaging, we can also record and measure qualitative difference in retinal imaging over drusen compared to age-matched non-AMD control eyes. The notable qualitative changes on SDOCT were the absence of the photoreceptor outer segments with the associated loss or disruption of the highly reflective photoreceptor inner segment/outer segment junction, the hyper-reflective haze within the remaining photoreceptor nuclear layer over drusen and hyper-reflective speckles over and adjacent to drusen. The findings of a decrease in photoreceptor height, focal loss of the photoreceptor outer segments and abnormal reflectivity at the site of remaining photoreceptors over drusen, supports the argument that neurosensory retinal degeneration is present in eyes with high risk drusen and is present before the development of geographic atrophy or choroidal neovascularization.
The hyper-reflective changes over drusen may represent multiple stages of a single process or potentially two different processes over drusen. Possible causes include an abnormal deposit such as pigment, displaced or migrated RPE cells or macrophages with pigment inside, exudates, lipid, or even blood within the retina or a degenerative retinal process. Some very intensely reflective intraretinal foci have been associated with hyperpigmentation on color fundus photographs.23
Focal pigment abnormalities have been associated with higher risk of progression to severe AMD,28, 30
and hyper-reflective sites on SDOCT may represent a precursor to larger pigment clumps, which are visible on fundus photographs, and thus might be an earlier risk indicator for druse progression. This question is being addressed in a longitudinal study ancillary to the Age Related Eye Disease Study 2 (AREDS2).31
In the AREDS2 Ancillary SDOCT Study (online at http://clinicaltrials.gov/ct2/show/NCT00345176
. Accessed October 6, 2008) larger SDOCT image datasets across the entire macula will be compared to conventional imaging in high risk drusen eyes with annual follow up for up to five years.
An inflammatory process or a degenerative retinal process such as in synaptic terminals could also cause focal PRL hyper-reflectivity if sub-cellular abnormalities increased the “normal” reflectivity of that site. We theorize that a degenerative cellular process may cause the less intense and more diffuse hyper-reflective haze visible within the photoreceptor nuclear layer. Because SDOCT reflectivity changes range from mild and diffuse to more focal and intense, these sites may represent a spectrum of abnormalities in photoreceptors overlying drusen. In postmortem studies, Johnson, et al. found that photoreceptors overlying and flanking drusen exhibited morphologic signs of degeneration including changes in localization of synaptic proteins within cells, the location and structure of the synaptic terminals, and in expression of stress response proteins.9, 10
Although abnormalities in the synaptic terminal of photoreceptor cells have also been shown in an experimental model of neovascular AMD, from which Caicedo et al. proposed that macrophages migrating into the neurosensory retina may initiate the neurosensory retinal dysfunction,32
a possible relationship between SDOCT hyper-reflectivity in the PRL and synaptic terminal dysfunction requires further study for validation.
In this SDOCT study, images of the PRL over drusen were captured at a single timepoint in the progression of AMD. Fortunately, with this non-contact method of imaging the living eye, we will be able to repeat these measurements over time to assess the relationships between drusen development and photoreceptor loss. To date, there are multiple theories regarding the relationship between drusen and the neurosensory retina in the early stages of AMD. These have ranged from photoreceptor degeneration (particularly rod loss) preceding further RPE loss and drusen progression8
to several theories of changes in Bruch’s membrane, drusen formation, or RPE degeneration causing photoreceptor degeneration.9,33
For example, drusen may affect photoreceptors by physical displacement damaging their structural integrity, or by compromising the function of RPE cells, a prerequisite for normal photoreceptor cell function and maintenance of the retinal microenvironment.9
Drusen may block the normal diffusion of metabolic materials between the photoreceptors and choroidal blood supply, leading to concentration of waste13,14
near the RPE and inhibition of the diffusion of oxygen, glucose, and nutritive serum-associated molecules required to maintain the health of the outer retina and RPE.13, 14
Although analysis of current SDOCT images will not aid in selecting between these theories, the use of these precise morphologic measurements over time, in conjunction with functional testing, should aid in the stratification of stages of disease progression. Moreover, since significant photoreceptor effects are already imaged with SDOCT in eyes with high-risk drusen, this imaging is also likely to be useful to investigate earlier events in drusen formation and photoreceptor change prior to high risk drusen. Detection and analysis of change in photoreceptors over smaller drusen in longitudinal studies might result in earlier predictors of AMD progression. Alternately, because with SDOCT we are now able to detect and measure changes in ultrastructural drusen components23
and the effect of drusen on the surrounding retina in vivo
, it is appropriate to search for either a relationship between these events and progression to severe disease or earlier signs of CNV or of geographic atrophy development.
The focal qualitative and quantitative changes in the PRL over drusen suggest measurable photoreceptor loss and likely dysfunction. These may be useful biomarkers for visual impairment associated with drusen and may predict the subsequent course of disease progression. Because the anomalous findings varied notably from patient to patient, there is a possibility that they are associated with specific genotype, systemic biomarkers, or with environmental exposure.
For SDOCT to be effective in AMD studies, researchers will need to know the reproducibility of qualitative grading, variance of drusen volume, photoreceptor, and other retinal layer measurements across a three-dimensional (3D) region of interest in the macula. This is in contrast to the more limited sample across the macula in this study. Investigators in the aforementioned AREDS2 Ancillary SDOCT Study will address these ideas in addition to the predictive value of qualitative and quantitative data in the longitudinal SDOCT imaging across a 3D central section of the macula.