A Systematic Comparison ofSpectral-Domain Optical CoherenceTomography and Fundus Autofluorescencein Patients with Geographic Atrophy

Ramzi G. Sayegh, MD,1 Christian Simader, MD,1 Ulrike Scheschy, MD,1 Alessio Montuoro, Dipl.Ing,1

Christopher Kiss, MD,1 Stefan Sacu, MD,1 David P. Kreil, PhD,2 Christian Prünte, MD,1

Ursula Schmidt-Erfurth, MD1

Purpose: To evaluate spectral-domain optical coherence tomography (SD-OCT) in providing reliable andreproducible parameters for grading geographic atrophy (GA) compared with fundus autofluorescence (FAF)images acquired by confocal scanning laser ophthalmoscopy (cSLO).

Design: Prospective observational study.Participants: A total of 81 eyes of 42 patients with GA.Methods: Patients with atrophic age-related macular degeneration (AMD) were enrolled on the basis of total

GA lesion size ranging from 0.5 to 7 disc areas and best-corrected visual acuity of at least 20/200. A novelcombined cSLO-SD-OCT system (Spectralis HRA-OCT, Heidelberg Engineering, Heidelberg, Germany) wasused to grade foveal involvement and to manually measure disease extent at the level of the outer neurosensorylayers and retinal pigment epithelium (RPE) at the site of GA lesions. Two readers of the Vienna Reading Centergraded all obtained volume stacks (20�20 degrees), and the results were correlated to FAF.

Main Outcome Measures: Choroidal signal enhancements and alterations of the RPE, external limitingmembrane (ELM), and outer plexiform layer by SD-OCT. These parameters were compared with the lesionmeasured with severely decreased FAF.

Results: Foveal involvement or sparing was definitely identified in 75 of 81 eyes based on SD-OCT by bothgraders (inter-grader agreement: ��0.6, P � 0.01). In FAF, inter-grader agreement regarding foveal involvementwas lower (48/81 eyes, inter-grader agreement: ��0.3, P � 0.01). Severely decreased FAF was measured overa mean area of 8.97 mm2 for grader 1 (G1) and 9.54 mm2 for grader 2 (G2), consistent with the mean SD-OCTquantification of the sub-RPE choroidal signal enhancement (8.9 mm2 [G1] �9.4 mm2 [G2]) and ELM loss with8.7 mm2 (G1) �10.2 mm2 (G2). In contrast, complete morphologic absence of the RPE layer by SD-OCT wassignificantly smaller than the GA size in FAF (R2�0.400). Inter-reader agreement was highest regarding completechoroidal signal enhancement (0.98) and ELM loss (0.98).

Conclusions: Absence of FAF in GA lesions is consistent with morphologic RPE loss or advanced RPEdisruption and is associated with alterations of the outer retinal layers as identified by SD-OCT. Lesion size isprecisely determinable by SD-OCT, and foveal involvement is more accurate by SD-OCT than by FAF.

Financial Disclosure(s): Proprietary or commercial disclosure may be found after the references.Ophthalmology 2011;118:1844–1851 © 2011 by the American Academy of Ophthalmology.

fsdafi

ttcp

Age-related macular degeneration (AMD) affects an esti-mated 30 to 50 million individuals worldwide1 and is themost common cause of legal blindness among elderly indi-viduals in developed countries.2,3 Despite the fact that aminority of patients affected by AMD have the neovascularform, no effective “treatment for the neovascular form” isavailable to reduce or reverse visual loss associated withatrophic AMD.4,5 Geographic atrophy (GA), a late stagedevelopment of AMD, occurs in 20% of patients withpreexisting clinical hallmarks of this degenerative dis-

ease.5–8 The natural course of GA, unlike the neovascular d

1844 © 2011 by the American Academy of OphthalmologyPublished by Elsevier Inc.

orm of AMD, progresses slowly and usually involves vi-ual loss, primarily due to retinal pigment epithelium (RPE)egeneration and neurosensory retinal atrophy resulting inbsolute and relative scotoma affecting the central visioneld.9–12

Retinal pigment epithelium cell death may result fromhe accumulation of lipofuscin, a toxic by-product of pho-oreceptor shedding, which accumulates in the lysosomalompartment of RPE cells.11,13 As a part of the agingrocess, the clearing ratio of lipofuscin in the RPE cell

iminishes and can impair normal cell metabolism. This

ISSN 0161-6420/11/$–see front matterdoi:10.1016/j.ophtha.2011.01.043

ciacamhlFBoPauRc

STI

IcOdfmgaptddwti

aa3aseTsarasBaeebfvsdasiro

Sayegh et al � SD-OCT and Fundus Autofluorescence in GA

eventually leads to cell death by induction of a reductivechronic inflammatory condition involving complement fac-tor H.14 Fundus autofluorescence (FAF) is predominantlybased on the fluorescent characteristics and distribution oflipofuscin within the RPE layer.13 Because metabolicallyaltered or lost RPE, as in dry AMD or GA, has differentfluorescent properties than physiologic RPE, autofluores-cence patterns could be used to identify pathophysiologicmechanisms in dry or atrophic AMD.15 Such changes in theFAF pattern of the RPE can be captured by a noninvasive invivo retinal imaging technique using confocal scanninglaser ophthalmoscopy (cSLO).13,16 The origins of atrophicprocesses in AMD and GA, however, are not entirely un-derstood.17,18

The evolution of optical coherence tomography (OCT)from time-domain OCT to spectral-domain OCT (SD-OCT)has greatly influenced AMD research and clinical knowl-edge.19,20 Currently, the benefit of OCT imaging particu-larly applies to a realistic analysis of the antiexudative effectof intravitreal antiangiogenic drugs in neovascularAMD.21–23

Compared with the diagnostic breakthrough in neovas-cular AMD, the role of OCT/SD-OCT with regard to GA inAMD is not clear. Several studies have compared FAF andSD-OCT with a specific focus on morphologic fea-tures.24–27 Only limited data are available on the potentialof SD-OCT to determine lesion size and the clinical repro-ducibility of planimetric measurements. Definite criteria toassess whether SD-OCT is indeed suitable for preciselylocalizing and delineating the lesion area have not yet beendefined, and the predictive value of SD-OCT features in GAprogression remains unclear.25,28 Because the alteration andloss of neurosensory elements at the retinal level seem todetermine visual prognosis in dry AMD,11 SD-OCT inpatients with GA seems essential if these changes are to befollowed and evaluated. The present study evaluates thecapacity of high-resolution, raster scanning SD-OCT toprovide dependable and reproducible parameters for grad-ing atrophic disease in comparison with FAF currently inuse as an anatomic end point parameter in several clinicaltrials. Such parameters would be of utmost value for diag-nosing and monitoring disease progression in patients withGA. Specific SD-OCT protocols would permit clinicians,investigators, and especially OCT reading centers to reli-ably evaluate the extent of preexisting lesions and the pro-gression rate of this late-stage dry AMD and objectivelyassess the efficacy of novel treatment strategies.

Patients and Methods

This prospective non-interventional case series was performed atthe Department of Ophthalmology at the Medical University ofVienna, Austria. The protocol and procedure followed the tenets ofthe Declaration of Helsinki. Written informed consent was ob-tained from each individual before inclusion in the study.

Inclusion/Exclusion Criteria and Follow-Up

Eighty-one eyes of 42 patients with severe vision loss due to GA

secondary to AMD were enrolled in the study and followed ac- l

ording to a standardized protocol in a prospective manner. Fornclusion in the study, patients had to be at least 55 years of agend to have no signs of choroidal neovascularization in biomicros-opy or OCT in either eye. If there was any doubt, fluoresceinngiography was performed. The macular condition was deter-ined by biomicroscopy, OCT, and FAF. The study eyes had to

ave a minimum GA lesion size of half a disc area and a maximumesion size of 7 disc areas, as estimated by biomicroscopy andAF, and the lesion had to fit in the SD-OCT volume scan.est-corrected visual acuity (BCVA) had to be �20/200. Clearcular media were required to provide good imaging quality.atients with history of any ocular disease that might confoundssessment of the retina were excluded. At each visit, patientsnderwent BCVA testing obtained using Early Treatment Diabeticetinopathy Study charts, slit-lamp examination, biomicroscopy,olor fundus photography, SD-OCT, and FAF.

pectral-Domain Optical Coherenceomography and Fundus Autofluorescence

maging Procedures

maging modalities analyzed in this study included SD-OCT and aSLO visualizing FAF, both integrated in the Spectralis HRA-CT (Heidelberg Engineering, Heidelberg, Germany). The cSLOevice uses blue light with excitation at 488 nm to illuminate theundus and detects emitted fluorescence signals from cellular ele-ents of retinal and RPE layers between 500 and 700 nm. A

rayscale FAF image with a frame size of 30�30 degrees wascquired in a high-speed mode with a resolution of 768�768ixels. For this specific study, non-normalized FAF data acquisi-ion was used and brightness and contrast were manually adjusteduring data acquisition for better visualization of the intensityistribution in the posterior pole. Fundus autofluorescence imagingas obtained by recording a 7-second video in the FAF mode of

he Spectralis HRA-OCT. One image comprising a mean FAFntensity was calculated out of 15 frames.

The SD-OCT device uses a superluminescence diode emittingscan beam at a wavelength of 870 nm. The retina is scanned atspeed of 40 000 A-scans per second with an axial resolution of

.9 �m and a transversal resolution of 14 �m. The eye tracker andutomatic real-time averaging modes of the Spectralis SD-OCTystem were used throughout the study. The eye tracker enablesach OCT scan to be registered and locked to a reference image.herefore, the Spectralis-OCT software can identify the previouscan location and scan the identical area. The automatic real-timeveraging mode, when activated, allows for adjustment of theecorded frames to obtain averaged B-scans, which enhances im-ge quality by reducing movement artifacts and optimizes theignal-to-noise ratio. The SD-OCT imaging protocol comprised 49-scans per volume scan of 20�20 degrees, and each scan wasveraged with 30 frames per B-scan. If the lesion size was notntirely displayed in the volume scan, the protocol allowed fornlarging the scan area. If acquisition of the scan was not possibleecause of the patient’s inability to comply, both the number oframes acquired for each B-scan and the number of B-scans in theolume scan were reduced. A minimum of 25 B-scans per volumecan with a minimum of 20 frames per B-scan were obtained. Theistance between the SD-OCT B-scans is 120 �m when acquiringvolume stack of 49 scans and 240 �m when acquiring a volume

tack of 25 B-scans. These distances are automatically registeredn the XML file generated by the Spectralis SD-OCT software. Theaster scan position for SD-OCT was centered in a manner anal-gous to the FAF imaging to delineate the complete extent of the

esion.

1845

Eih(cci

IAO

FamlSp

d2agawFfWp

S

DmpTBgcqfdCtaa

pcboliaw

arlbotn

Ophthalmology Volume 118, Number 9, September 2011

Planimetric Measurements of GeographicAtrophy Lesions by Spectral-DomainOptical Coherence Tomography

To planimetrically measure areas in SD-OCT scans and subse-quently compare them with FAF, we developed a program (OCT-Tool Kit) that is able to read and display data generated by theSpectralis XML Data Export Interface. Marking an area of interestis performed by manually drawing a line in the B-scan image andprojecting it to the en face image. The mean distance from 1B-scan to the next was used as the width of the line whenprojecting it to the en face image. The software uses the calibration(XML file) and the manually marked data to calculate the size ofthe areas marked in the B-scans. Defined neurosensory retinal andRPE layers were measured individually and included choroidalsignal enhancement below the RPE layer, the presence or absenceof the RPE cell layer, the outer plexiform layer (OPL), and theexternal limiting membrane (ELM). The nomenclature of the ret-inal layers in the SD-OCT B-scans is in accordance with Schmidt-Erfurth et al.19

Comparison of Geographic AtrophyFeatures Between Spectral-Domain OpticalCoherence Tomography and FundusAutofluorescence

To correlate SD-OCT with FAF findings, the same 2 independentreaders from the Vienna Reading Center (VRC) measuring thelesion size in FAF separately graded the obtained SD-OCT B-scanbaseline data obtained from the entire lesion area using the OCTplanimetric protocol according to the areas of interest: For eachOCT B-scan, the complete and questionable presence of choroidalsignal enhancement, referred to as area 1 and area 2, respectively,were graded. Complete choroidal signal enhancement was definedas a consistent signal enhancement throughout the choroidea thatcan be definitively determined by the reader; questionable choroi-dal signal enhancement was defined as choroidal signal enhance-ment that cannot be graded as choroidal signal enhancement withcertainty. Furthermore, the sum of both areas was referred to asarea 1/2 (Figs 1–3). Complete loss of the RPE was defined as area3, incomplete loss of the RPE layer was defined as area 4, andcomplete loss of both areas was defined as area 3/4 (Figs 1–3).Complete loss of the RPE was defined as a complete reduction ofthe RPE layer to a thin line with no hyperreflective elements in theRPE layer. Incomplete loss of the RPE was defined as a partialreduction of the RPE layer when compared with normal RPEappearance.

With regard to atrophy of the neurosensory layers of the retina,a thinning and shifting of the OPL toward the RPE layer waslabeled as area 5, and loss of the OPL was defined as area 6. Lossof the ELM was defined as area 7 (Fig 2).

The OCT-Tool Kit summarizes the measurements and indicatesplanimetric values for each specific area of interest, which can thenbe compared and correlated to the FAF measurement values. Thedefined areas demonstrating GA features were then correlated withBCVA.

Evaluation of Lesion Size by FundusAutofluorescence

The GA lesion size in the FAF images was measured manually by

2 readers from the VRC using the Spectralis software “Heidelberg r

1846

ye Explorer” region overlay device. Fundus autofluorescencemages were enlarged, and the GA area, represented as an area ofypofluorescence in FAF, was delineated using an external stylusWacom Bamboo Pen & Touch; Wacom Technology Corp., Van-ouver, WA). Those results served as reference values whenomparing the dimension of GA in FAF with the areas of interestn SD-OCT B-scans.

dentification of Foveal Sparing in Fundusutofluorescence and Spectral-Domainptical Coherence Tomography

urthermore, we examined whether SD-OCT and FAF were equiv-lent with respect to identifying foveal sparing or foveal involve-ent in the GA process. An adequate grading system was estab-

ished, and the 2 readers from the VRC independently graded theD-OCT B-scans and FAF images at baseline according to thisredetermined grading system.

The grading system was established according to the level ofisease progression: grade 1 � fovea definitely involved; grade� fovea probably involved; grade 3 � fovea probably spared;

nd grade 4 � fovea definitely spared. In SD-OCT, the fovea wasraded as fovea definitely involved when the RPE in the fovea wasltered causing choroidal signal enhancement. In FAF, the foveaas defined as definitely involved when the grayscale image inAF had similar intensity values in the region of the supposedovea and in the region of GA in the rest of the posterior pole.

hen in doubt, the graders were asked to grade the fovea asrobably involved or spared.

tatistical Analysis

ata analysis was performed in an established statistical environ-ent (R version 2.9.2). First-order descriptive statistics were com-

uted to provide a simple characterization of the measured values.o relate grader classifications of SD-OCT or FAF images toCVA values, a linear logistic model was fit, also testing for arader effect and any left/right bias. Left/right bias was insignifi-ant for all tests and was thus removed from the models. Touantify inter-grader concordance concerning the grading of theoveal involvement, � coefficients for the reliability of nominalata were computed. The original � coefficient as computed byohen29 and the �= coefficient for pooled classification propor-

ions30 were similar and significant with comparable P values. Inddition, we report the �� coefficient adjusted for prevalenceccording to Byrt et al.31

Linear regression analysis was used to examine alternativeotential relationships of the different areas of interest in SD-OCTompared with FAF. Non-transformed measurements were used,ecause they yielded better model fits than models of square-rootr log-transformed data. In multivariate analysis of variance, theeast significant factors were trimmed iteratively until all remain-ng factors were significant (P � 0.05). Univariate regressionchieved only a marginally worse model fit; therefore, the focusas set on reporting the best univariate models.

For the correlation of graded GA areas with BCVA, all visualcuity values were converted to logarithm of minimal angle ofesolution for statistical analysis. Univariate models on linear andog scales, with arbitrary and zero offsets, were considered. Theest model fit for area 3 was achieved on a linear scale with zeroffset, and for all other variables, best model fits were achieved onhe log scale with zero offset, that is, by fitting a power law witho scaling factor. Correlation values are each given for the linear

egression fit of the linear or log-transformed data.

Ftdmapexternal limiting membrane (ELM).

Sayegh et al � SD-OCT and Fundus Autofluorescence in GA

Results

Included Study EyesEighty-one eyes of 42 patients (16 men and 26 women; mean age 80years, range 55–89 years) were included in this prospective study. Allpatients presented bilateral GA. Of the potential total number of 84study eyes, 3 fellow eyes were excluded because of poor retinalimaging quality due to advanced cataract. Of the 81 included eyes, 41were right eyes and 40 were left eyes. Mean BCVA at baseline amongthe 81 study eyes was 20/51 Snellen equivalents.

Determination of Foveal Sparing bySpectral-Domain Optical CoherenceTomography and Fundus Autofluorescence, andCorrelation with Best-Corrected Visual AcuityFor all 81 eyes, foveal involvement by the atrophic process wasindependently assessed from SD-OCT and FAF images. Gradingresults for both imaging methods are summarized in Table 1 (avail-able at http://aaojournal.org), and examples are shown in Figure 4.

Logistic regression indicated a highly significant relation between theBCVA and the grading of SD-OCT measurements (P � 2�10�8). Thisfinding shows that clear results could be obtained by SD-OCT.

Independence of the results of the operators was tested by including agrader effect in the logistic regression model. There was no significantgrader bias (P � 0.54) and little random variation (���’�0.6, P � 0.01;�’’�0.9). This is reflected in the high inter-individual agreement betweengraders (Table 2, available at http://aaojournal.org), where congruentSD-OCT gradings were obtained for 75 of 81 eyes: grade 1 (definiteinvolvement) for 70 eyes and grade 4 (definite sparing) for 5 eyes. Inthese groups, mean BCVA values were 20/61 and 20/26, respectively. In6 eyes, graders were discordant regarding involvement or sparing of thefovea by the atrophic process. Mean BCVA was 20/27 in this group.

Figure 1. Graded area of hypofluorescence measured in fundus autofluo-rescence (FAF) representing the area of geographic atrophy (GA).

igure 2. Areas of interest graded in spectral-domain optical coherenceomography (SD-OCT) (from top to bottom): area 1 � complete choroi-al signal enhancement; area 2 � questionable choroidal signal enhance-ent; area 3 � complete absence of retinal pigment epithelium (RPE);

rea 4 � incomplete loss of RPE; area 5 � subsidence of the outerlexiform layer (OPL); area 6 � loss of the OPL; area 7 � loss of the

1847

Fw

tlRctcwahcoo

CSV

Wbfvv(Efwpimw

Ophthalmology Volume 118, Number 9, September 2011

The BCVA was significantly related to grading of FAF mea-surements (P � 5�10�4). For FAF, the grader effect in the logis-tic regression model was marginally significant (P � 0.08), sug-gesting a possible operator bias. Little random variation wasobserved (���’�0.3, P � 0.01; �’’�0.4). Inter-individual agree-ment between graders, with consistent FAF gradings, was obtainedfor 48 of 81 eyes. In 40 eyes, definite foveal involvement (grade 1)was identified, with a mean BCVA of 20/74, whereas in 5 eyes thefovea was identified as probably involved (grade 2, BCVA �20/38). Finally, in 3 eyes the graders considered the fovea to beprobably spared (grade 3, mean BCVA of 20/27). For the remain-ing 33 eyes, there was inter-grader disagreement: grade 1 versusgrade 2 for 18 eyes (mean BCVA � 20/42), grade 1 versus grade3 for 8 eyes (BCVA � 20/77), and grade 2 versus grade 3 for 7eyes (BCVA � 20/29). In general, there was an inconclusiverelationship between the FAF-based identification of foveal in-volvement and BCVA within each group and between groups(Table 2, available at http://aaojournal.org).

Determination of Geographic Atrophy Lesion SizeBased on Spectral-Domain Optical CoherenceTomography Grading and FundusAutofluorescence Planimetry

The results of the evaluation of GA features in absolute planim-etric values (Table 3, available at http://aaojournal.org) were sim-ilar for both graders regarding complete choroidal signal enhance-ment through increased transmission of light in the absence ofmelanin (area 1),32 subsidence of the OPL (area 5), loss of theELM (area 7) by SD-OCT, and absence of FAF by cSLO. Lesionsize values referring to complete loss of RPE (area 3) wereapproximately half the dimension of the areas mentioned above,including absence of FAF. For areas 3 and 4, comprising complete

Figure 3. Combined areas of interest in spectral-domain optical coherencand questionable choroidal signal enhancement. B, Combined areas 3 anincomplete loss of RPE.

and incomplete RPE loss from SD-OCT measurements, similar a

1848

AF results indicating GA with complete loss of RPE fluorescenceere obtained.

Complete choroidal signal enhancement was probably closesto representing the area of GA in SD-OCT B-scans measured in aarger area than complete RPE loss or advanced alteration of thePE based on RPE morphology in SD-OCT. The total extent ofomplete and incomplete RPE loss, however, was consistent withhe size of the area of complete choroidal signal enhancement andomplete loss of FAF. The areas of ELM loss and OPL shiftingere consistent with the sum of the areas of complete RPE loss

nd advanced alteration of RPE morphology (Table 3, available atttp://aaojournal.org). In all eyes, both graders were able to pre-isely measure a larger area of complete (area 1) and a smaller areaf questionable choroidal signal enhancement of incomplete lossf RPE and loss of the ELM.

omparison of Fundus Autofluorescence andpectral-Domain Optical Coherence Tomographyalues

e tested a multivariate model relating the area measured in FAFy the graders to the areas graded in SD-OCT. The least significantactors were trimmed iteratively until all remaining analysis ofariance factors were significant (P � 0.05). The obtained multi-ariate model explaining FAF as a function of the graded area 1complete choroidal signal enhancement) and area 7 (loss of theLM) achieved an adjusted R2�0.961. The best univariate model

or the FAF was with area 1, obtaining a comparable fit qualityith an adjusted R2�0.958. These findings justify focusing on theresentation of the correlations corresponding to multiple univar-ate models (Table 4, available at http://aaojournal.org). Theatching scatter plots comparing the results of SD-OCT gradingsith the FAF are shown in Figure 5. Although all areas exhibited

ography (SD-OCT). A, Combined areas 1 and 2 representing completeepresenting complete absence of retinal pigment epithelium (RPE) and

e tomd 4 r

significant non-zero correlation, only the graded area 1 (complete

c(

CFOBTi

D

TpS

d.

Sayegh et al � SD-OCT and Fundus Autofluorescence in GA

choroidal signal enhancement), area 5 (OPL shifting), and area 7(loss of the ELM) in SD-OCT had a strong correlation with thehypofluorescent area measured in FAF. In particular, area 1 rep-resenting the area of complete choroidal signal enhancement in theSD-OCT B-scans correlated best and was also less dependant ongrader variability (data not shown).

The area of complete choroidal signal enhancement in SD-OCTB-scans correlated strongly with the area of OPL shifting and withELM loss. Moreover, the sum of the areas of complete andquestionable choroidal signal enhancement correlated highly withthe sum of the areas of complete and incomplete RPE loss (Table5, available at http://aaojournal.org).

Inter-grader Reproducibility of GradingMeasurements

The inter-grader reproducibility of the FAF and SD-OCT measure-ments of lesion size were best for FAF, followed closely by

Figure 4. Grading of the foveal involvement of the atrophic process in fun(SD-OCT). A, In FAF fovea probably involved, in SD-OCT fovea definiteC, In FAF fovea definitely involved, in SD-OCT fovea definitely involve

Figure 5. Agreement between area of hypofluorescence measured in fundusgraded areas of interest in spectral-domain optical coherence tomography (SD3 represents the area of complete absence of retinal pigment epithelium (RPE

area of outer plexiform layer (OPL) subsidence, and area 7 represents the area of los

omplete choroidal signal enhancement and loss of the ELMTable 6 and Fig 6, available at http://aaojournal.org).

orrelation of Best-Corrected Visual Acuity,undus Autofluorescence, and Spectral-Domainptical Coherence Tomography Lesion Sizeetween the Two Gradershe BCVA correlated significantly with all graded areas as shown

n Table 7 (available at http://aaojournal.org).

iscussion

his prospective study represents the first analysis of com-lete SD-OCT volume stacks to determine the potential ofD-OCT versus FAF for monitoring patients with GA, partic-

tofluorescence (FAF) and spectral-domain optical coherence tomographyred. B, In FAF fovea probably spared, in SD-OCT fovea definitely spared.

orescence (FAF) representing the area of geographic atrophy (GA) and the). Area 1 represents the area of complete choroidal signal enhancement, areaa 4 (A4) represents the area of incomplete loss of RPE, area 5 represents the

dus auly spa

autoflu-OCT), Are

s of the external limiting membrane (ELM). FAF � fundus autofluorescence.

1849

rtialtttrtpnlflboaa

cibavoRtm

aGrpemf

R

Ophthalmology Volume 118, Number 9, September 2011

ularly with respect to identifying the foveal condition andrelevant grading parameters for precisely measuring GA.

Lujan et al28 compared the size of GA in SD-OCT andFAF in 5 eyes by superimposing both imaging modalitiesand delineating the borders of the disease without analyzingthe retinal layer morphology. The findings indicated that itis in principle possible to assess the size of an atrophic GAlesion in SD-OCT, but retinal morphology was not an issue.

Choroidal signal enhancements and alterations at thelevel of RPE, ELM, and OPL are characteristic morphologicchanges in GA that we assumed might be helpful in accuratelymeasuring the lesion size. The aim of our study was to deter-mine which of these pathologic changes should be graded toobtain equivalent planimetric measurements in FAF.

Although general grading reproducibility was good for bothmethods, foveal sparing was identified by SD-OCT with ahigher certainty and inter-grader agreement than with FAF.This could be due to the presence of yellow macular pigmentin the neurosensory retina at the fovea that blocks the blueexcitation light of FAF imaging. Moreover, the grayscale im-ages obtained by FAF merely reflect the overall autofluores-cence of the RPE and the quantity of fluorophores within theRPE cells, if present. Neither the resolution nor the quantifi-cation of FAF values allow for precise delineation regardingthe parameter “foveal localization,” and the fovea is frequentlylocated within the junctional zone of RPE disease, which, asour study also shows, is generally problematic for FAF eval-uation. Furthermore, foveal depression can easily be deter-mined by SD-OCT. Therefore, screening for decreased FAF inthe central fovea may be difficult without complementaryextensions, such as near-infrared reflectance imaging. Conclu-sions regarding foveal involvement or sparing based on FAFalone must therefore be drawn cautiously.27

The agreement of outcomes between grader 1 and grader 2regarding the parameters graded by SD-OCT highlight thefeasibility of objectively grading well-defined parameters inSD-OCT. A significant correlation of these SD-OCT–basedresults with the GA area measured in FAF proves that bothmethods can be used to consistently quantify the extent of thisdisease.

There were significant correlations of varying strengthbetween the area of interest and the visual acuity, which canbe explained by the fact that the size of the GA does notentirely determine the BCVA in GA, because there is oftenfoveal sparing. As the GA area expands, the probability offoveal involvement increases.17,27 Furthermore, the resultsobtained and presented in Table 1 (available at http://aao-journal.org) show that mean BCVA among patients withdefinite fovea involvement ranged between 20/57 and 20/77. This indicates that BCVA in patients with central GAmay be higher than generally assumed.

The results of our study showed that the definite RPEloss graded in SD-OCT and FAF was only weakly corre-lated (R2�40%; Table 4, available at http://aaojournal.org).This correlation increased to 97% when we added the areasof definite RPE loss to that of uncertain RPE alteration withmoderate RPE loss. We presume that this issue underlinesthe importance of the junctional zone in GA because thiszone is the site of disease activity and will be the target for

any effective therapeutic strategy. The interest in this area is

1850

eflected by the large number of scientific publica-ions.25,26,33,34 In the study by Bearelly et al,25 SD-OCTmages of the junction zone in patients with GA werenalyzed to determine whether RPE loss or photoreceptoross is the first sign of GA progression. The authors statedhat if this were the case, then the frequency of photorecep-or loss outside the GA margin would have been higher. Inhe present study, we analyzed the alterations of the photo-eceptor layer in complete B-scans and concluded that pho-oreceptor alterations seem to be more extensive than com-lete RPE alterations. A longitudinal study in this field isecessary to provide more detailed insight, and the lack ofongitudinal data is a limitation of this study. Fundus auto-uorescence does not allow for accurate mapping of the fineorders of a disease that is characterized by different stagesf progression and cannot differentiate between an absent ordiseased RPE in the same manner as SD-OCT, which

llows for viewing and grading all retinal layers.Most noteworthy is the finding that not only RPE absence

orrelates closely with neurosensory alterations, but that evenn junctional zones where RPE cells are still present but areeginning to undergo morphologic changes, retinal layers suchs the ELM and OPL are severely affected. It remains contro-ersial whether the primary origin of atrophic AMD is the RPEr neuronal elements such as photoreceptors.35,36 The fact thatPE atrophy occurs at the novel macular site after 360-degree

ranslocation36 suggests that photoreceptor disease is the pri-ary stimulus for subsequent RPE death.In conclusion, spectral-domain OCT seems to be an

ppropriate imaging modality for evaluating the extent ofA lesions. The retinal scanning time in SD-OCT and FAF

elevant to obtaining gradable material is dependent on theatient and physician. Manual grading of the B-scans, how-ver, is time-consuming and should be replaced by auto-ated algorithms delineating choroidal signal enhancement

or accurate GA lesion size determination.

eferences

1. Gehrs KM, Anderson DH, Johnson LV, Hageman GS. Age-related macular degeneration—emerging pathogenetic andtherapeutic concepts. Ann Med 2006;38:450–71.

2. Resnikoff S, Pascolini D, Etya’ale D, et al. Global data onvisual impairment in the year 2002. Bull World Health Organ2004;82:844–51.

3. Bressler NM, Bressler SB, Fine SL. Age-related maculardegeneration. Surv Ophthalmol 1988;32:375–413.

4. Rosenfeld PJ, Brown DM, Heier JS, et al, MARINA StudyGroup. Ranibizumab for neovascular age-related macular de-generation. N Engl J Med 2006;355:1419–31.

5. Klein R, Klein BE, Knudtson MD, et al. Fifteen-year cumu-lative incidence of age-related macular degeneration: the Bea-ver Dam Eye Study. Ophthalmology 2007;114:253–62.

6. Ferris FL III, Fine SL, Hyman L. Age-related macular degen-eration and blindness due to neovascular maculopathy. ArchOphthalmol 1984;102:1640–2.

7. Hyman LG, Lilienfeld AM, Ferris FL III, Fine SL. Senilemacular degeneration: a case-control study. Am J Epidemiol

1983;118:213–27.

2

2

2

2

2

2

2

3

3

3

3

3

3

3

Sayegh et al � SD-OCT and Fundus Autofluorescence in GA

8. Klein R, Klein BE, Lee KE, et al. Changes in visual acuity ina population over a 15-year period: the Beaver Dam EyeStudy. Am J Ophthalmol 2006;142:539–49.

9. Maguire P, Vine AK. Geographic atrophy of the retinal pig-ment epithelium. Am J Ophthalmol 1986;102:621–5.

10. Schatz H, McDonald HR. Atrophic macular degeneration: rateof spread of geographic atrophy and visual loss. Ophthalmol-ogy 1989;96:1541–51.

11. Green WR, Key SN III. Senile macular degeneration: a his-topathologic study. Trans Am Ophthalmol Soc 1977;75:180–254.

12. Sunness JS. The natural history of geographic atrophy, theadvanced atrophic form of age-related macular degenera-tion. Mol Vis [serial online] 1999;5:25. Available at:http://www.molvis.org/molvis/v5/a25/. Accessed December16, 2010.

13. Delori FC, Dorey CK, Staurenghi G, et al. In vivo fluores-cence of the ocular fundus exhibits retinal pigment epitheliumlipofuscin characteristics. Invest Ophthalmol Vis Sci 1995;36:718–29.

14. Okemefuna AI, Nan R, Miller A, et al. Complement factor Hbinds at two independent sites to C-reactive protein in acutephase concentrations. J Biol Chem 2010;285:1053–65.

15. Bindewald A, Bird AC, Dandekar SS, et al. Classification offundus autofluorescence patterns in early age-related maculardisease. Invest Ophthalmol Vis Sci 2005;46:3309–14.

16. von Ruckmann A, Fitzke FW, Bird AC. Fundus autofluores-cence in age-related macular disease imaged with a laserscanning ophthalmoscope. Invest Ophthalmol Vis Sci 1997;38:478–86.

17. Sunness JS, Gonzalez-Baron J, Applegate CA, et al. Enlarge-ment of atrophy and visual acuity loss in the geographicatrophy form of age-related macular degeneration. Ophthal-mology 1999;106:1768–79.

18. Dreyhaupt J, Mansmann U, Pritsch M, et al. Modelling thenatural history of geographic atrophy in patients with age-related macular degeneration. Ophthalmic Epidemiol 2005;12:353–62.

19. Schmidt-Erfurth U, Leitgeb RA, Michels S, et al. Three-dimensional ultrahigh-resolution optical coherence tomogra-phy of macular diseases. Invest Ophthalmol Vis Sci 2005;46:3393–402.

20. Cukras C, Wang YD, Meyerle CB, et al. Optical coherencetomography-based decision making in exudative age-relatedmacular degeneration: comparison of time- vs spectral-domaindevices. Eye (Lond) 2010;24:775–83.

21. Witkin AJ, Vuong LN, Srinivasan VJ, et al. High-speed ul-trahigh resolution optical coherence tomography before andafter ranibizumab for age-related macular degeneration. Oph-thalmology 2009;116:956–63.

22. Sayanagi K, Sharma S, Yamamoto T, Kaiser PK. Comparison

of spectral-domain versus time-domain optical coherence to-

Footnotes and Financial Disclosures

The author(s) have made the following disclosure(s): Christian Prünte,

MBp

P2

CCvE

mography in management of age-related macular degenerationwith ranibizumab. Ophthalmology 2009;116:947–55.

3. Kiss CG, Geitzenauer W, Simader C, et al. Evaluation ofranibizumab-induced changes in high-resolution optical co-herence tomographic retinal morphology and their impact onvisual function. Invest Ophthalmol Vis Sci 2009;50:2376–83.

4. Wolf-Schnurrbusch UE, Enzmann V, Brinkmann CK, Wolf S.Morphologic changes in patients with geographic atrophyassessed with a novel spectral OCT-SLO combination. InvestOphthalmol Vis Sci 2008;49:3095–9.

5. Bearelly S, Chau FY, Koreishi A, et al. Spectral domainoptical coherence tomography imaging of geographic atrophymargins. Ophthalmology 2009;116:1762–9.

6. Brar M, Kozak I, Cheng L, et al. Correlation between spectral-domain optical coherence tomography and fundus autofluo-rescence at the margins of geographic atrophy. Am J Ophthal-mol 2009;148:439–44.

7. Schmitz-Valckenberg S, Fleckenstein M, Helb HM, et al. Invivo imaging of foveal sparing in geographic atrophy second-ary to age-related macular degeneration. Invest OphthalmolVis Sci 2009;50:3915–21.

8. Lujan BJ, Rosenfeld PJ, Gregori G, et al. Spectral domainoptical coherence tomographic imaging of geographic atro-phy. Ophthalmic Surg Lasers Imaging 2009;40:96–101.

9. Cohen J. A coefficient of agreement for nominal scales. EducPsychol Meas 1960;20:37–46.

0. Siegel S, Castellan NJ Jr. Nominally Scaled data and theKappa Statistics K. In: Anker JD, ed. Nonparametric Statisticsfor the Behavioral Sciences. 2nd ed. New York: McGraw-Hill;1988:284–91.

1. Byrt T, Bishop J, Carlin JB. Bias, prevalence and kappa. J ClinEpidemiol 1993;46:423–9.

2. Unterhuber A, Povazay B, Hermann B, et al. In vivo retinaloptical coherence tomography at 1040 nm - enhanced pene-tration into the choroid. Opt Express [serial online] 2005;13:3252–8. Available at: http://www.opticsinfobase.org/abstract.cfm?URI�oe-13-9-3252. Accessed December 16, 2010.

3. Schmitz-Valckenberg S, Bultmann S, Dreyhaupt J, et al. Fun-dus autofluorescence and fundus perimetry in the junctionalzone of geographic atrophy in patients with age-related mac-ular degeneration. Invest Ophthalmol Vis Sci 2004;45:4470–6.

4. Bindewald A, Schmitz-Valckenberg S, Jorzik JJ, et al. Clas-sification of abnormal fundus autofluorescence patterns in thejunctional zone of geographic atrophy in patients with agerelated macular degeneration. Br J Ophthalmol 2005;89:874–8.

5. Sahel JA. Saving cone cells in hereditary rod diseases: apossible role for rod-derived cone viability factor (RdCVF)therapy. Retina 2005;25(Suppl):S38–9.

6. Eckardt C, Eckardt U. Macular translocation in nonexudative

age-related macular degeneration. Retina 2002;22:786–94.

Originally received: January 8, 2010.Final revision: January 13, 2011.Accepted: January 13, 2011.Available online: April 15, 2011. Manuscript no. 2010-46.1 Department of Ophthalmology, Medical University of Vienna, Austria.2 Chair of Bioinformatics, Department of Biotechnology, Boku UniversityVienna, Austria.

Financial Disclosure(s):

D, has a financial relationship with Novartis Pharma, Alcon Pharma, andayer. None of the authors have a proprietary interest in any of theroducts mentioned in this study.

arts of the study were presented at: the DOG annual meeting, September6, 2009, Leipzig, Germany (paper presentation)

orrespondence:hristian Simader, MD, Department of Ophthalmology, Medical Uni-ersity of Vienna, Austria, Waehringer Guertel 18-20, Vienna, Austria.

-mail: chrisitan.simader@meduniwien.ac.at

1851