RETINAL THICKNESS IN DIABETICRETINOPATHYA Study Using Optical Coherence Tomography(OCT)

WINFRIED GOEBEL, MD, TATJANA KRETZCHMAR-GROSS, MD

Background and Objective: The authors conducted a controlled study to quantifymacular retinal thickness in diabetic retinopathy using optical coherence tomography(OCT) as an objective and noninvasive tool. The relationship between retinal thickness andstandard methods of evaluating macular edema was investigated.

Patients and Methods: A total of 136 patients in different stages of diabetic retinopathywere examined with OCT. In addition, fluorescein angiograms as well as standard eyeexaminations were conducted. The control group consisted of 30 individuals with a normalmacula.

Results: In the controls, retinal thickness was 153 ! 15 !m in the fovea, 249 ! 19 !min the temporal parafoveal region, and 268 ! 20 !m in the nasal parafoveal region. Indiabetic patients, retinal thickness was increased to 307 ! 136 !m in the fovea, 337 ! 88!m in the temporal retina, and 353 ! 95 !m in the nasal retina, respectively. Thedifferences between diabetics and controls were highly significant (P " 0.001). Retinalthickening correlated with fluorescein leakage in the angiograms to some extent. Therewas an intermediate correlation between retinal thickness and visual acuity, particularly inpatients without macular ischemia. Sensitivity of detecting clinically significant macularedema by measuring foveal retinal thickness was 89% and specificity was 96%.

Conclusion: Optical coherence tomography allows us to quantify retinal thickness indiabetic retinopathy with excellent reproducibility. OCT is able to detect sight-threateningmacular edema with great reliability.

RETINA 22:759–767, 2002

D iabetic macular edema is one of the main causesof visual loss in diabetic patients.1 So far, diabetic

maculopathy has been routinely graded by slit-lampbiomicroscopy, fundus photography, and fluorescein an-giography.2 The interpretation of examination results,however, is subjective and lacks inter- and intraobserver

reproducibility.3 In the course of multicenter trials, grad-ing systems for fundus photographs4 as well as angio-grams5 have been developed. Although such gradingstrategies provide some means of semiquantitative as-sessment of diabetic maculopathy, they are relativelycomplicated, time consuming, and difficult to use in thedaily routine. In the past few years, new imaging tech-niques that have been used to quantify macular edemahave emerged.6,7There are several phenomena that relate to macular

edema, with breakdown of the blood–retinal barrierand increase in retinal volume and hence retinal thick-

From the Department of Ophthalmology, University of Wuerz-burg, Germany.The authors have no financial or proprietary interest in the

instrumentation used.Reprint requests: Winfried Goebel, MD, Department of Oph-

thalmology, University of Wuerzburg, Josef-Schneider-Str. 11,97080 Wuerzburg, Germany.

759

ness being the most important parameters.6 As a re-finement of fluorescein angiographic and fluoropho-tometric techniques, the retinal leakage analyzer(RLA), a prototype instrument, permits the quantita-tive analysis of the blood–retinal barrier in combina-tion with topographic imaging of the retina.8 With theRLA, the permeability of the blood–retinal barrier wasshown to be increased not only in manifest but also insubclinical macular edema.9 To some degree, break-down of the blood–retinal barrier and increase inretinal volume may be independent of each otherbecause fluid accumulation in the retina can occur notonly from vasogenic causes and a disturbed equilib-rium between fluid extravasation and reabsorption, butalso from toxic effects.7,9,10 Therefore, analysis of theblood–retinal barrier alone, e.g., with fluorescein an-giography or fluorophotometry, may not be sufficientto fully appreciate macular edema. Three-dimensionaltopography of the macula with scanning laser ophthal-moscopes, e.g., the Heidelberg Retina Tomograph(HRT), provides indirect measurement of retinal vol-ume and has demonstrated an increase in the meanvolume above an assumed reference plane in patientswith clinically detectable macular edema.11 Hudsonand coworkers12 proposed a new technique for theanalysis of scanning laser tomography images. In asmall series, they evaluated the Z-profile signal widthof laser tomography images and were able to demon-strate that the Z-profile signal width increases in dia-betic macular edema. Although the reproducibility ofthis technique is acceptable, the depth resolution islimited (around 150 !m) and the definition of thereference plane may be a source of error.13The most sensitive parameter that relates to macular

edema seems to be retinal thickness. Macular thick-ness can be measured using the retinal thickness an-alyzer (RTA), an instrument that computes retinalthickness from oblique laser slit projections on theposterior pole.10,14 Retinal thickness was shown to beincreased in patients with diabetic retinopathy com-pared to controls.15–17 Although the technical preci-sion of this technology is approximately 50 !m, itmay be degraded by opacities of the optical media orintraretinal lesions common in diabetic retinopathy,such as hemorrhages or exudates.Optical coherence tomography (OCT) is another

new, noninvasive technique that provides retinalthickness profiles of the central retina with a technicalresolution of 10 !m to 20 !m; this technique permitsan objective assessment of retinal thickness indepen-dent of the refractive status of the eye.18 Severalstudies19–21 have demonstrated that OCT is able tomeasure retinal thickness in diabetic macular edemaobjectively and with high accuracy. Intraretinal

changes can be clearly separated from the reflection ofthe pigment epithelium/choriocapillaris complex. Be-cause of the longer acquisition time, eye movementsmay have a greater influence on the measurementscompared to the RTA.Among the new imaging techniques, OCT seems to

be most promising in the clinical analysis of diabeticmacular edema. We therefore measured macular reti-nal thickness with OCT in a cross-sectional study ofpatients with diabetic maculopathy and investigatedthe relationship between retinal thickness and standardmethods of evaluating macular edema, visual acuity,slit-lamp biomicroscopy, and fluorescein angiography.Specificity and sensitivity of detecting diabetic macularedema by measuring retinal thickness are reported.

Materials and Methods

Patients

A total of 150 consecutive patients with diabeteswere recruited from our outpatient department over 8months. Inclusion criteria were diabetic retinopathy ofany stage in at least one eye and absence of otherretinal diseases (e.g., macular degeneration, retinalvein occlusion, hypertensive retinopathy). Exclusioncriteria were vitreous hemorrhage, advanced cataract,or significant corneal opacities, as well as intraocularinflammation. Adequate visibility of fundus detail wasrequired for the study. Photocoagulation (focal or pan-retinal) less than 4 months prior or intraocular surgeryother than cataract extraction also precluded enroll-ment in the study.Informed consent was obtained before inclusion

and the study was performed in accordance with com-mon ethical standards. A standard eye examinationincluding lens status, refraction (Autorefractor CanonR-30), best-corrected visual acuity (Snellen equiva-lent), and slit-lamp biomicroscopy (90-diopter [D]lens) of the fundus was performed. The presence ofclinically significant macular edema (CSME) accord-ing to Early Treatment of Diabetic Retinopathy Study(ETDRS) standards22 was noted. When the presenceof CSME was questionable during examination withthe 90-D precorneal lens, a Goldman contact lens wasused in addition to confirm the diagnosis. The bloodlevel of glycosylated hemoglobin (HBA1c) was deter-mined to assess the quality of blood sugar control.Thirty consecutive subjects with normal macula and

central retina on biomicroscopy served as a controlsample. Individuals were only eligible in the normalgroup if they had no history of retinal diseases, pri-mary or secondary glaucoma, intraocular inflamma-tion, intraocular surgery apart from cataract extraction

760 RETINA, THE JOURNAL OF RETINAL AND VITREOUS DISEASES ● 2002 ● VOLUME 22 ● NUMBER 6

more than 4 months prior, diabetes mellitus, or un-controlled arterial hypertension.

Optical Coherence Tomography

We used the commercially available OCT unit fromZeiss-Humphrey (OCT 2000 Scanner, Zeiss-Hum-phrey, San Leandro, CA), Software Revision A4. Itcombines a scanning low coherence interferometerworking at an 840-!m wavelength with a video cam-era to provide fundus visualization. Each cross-sec-tional scan consists of 100 longitudinal A-scans. In-traretinal reflectivity in the optical cross-section isoutput in false-color representation. The techniquewas introduced into ophthalmology in 1993 by Swan-son et al23 and has been described in detail by Hee andcoworkers.24Every patient had four standardized OCT scans for

each eye: horizontal scans through the fovea with adiameter of 2.8 mm repeated three times and a verticalscan of the same length through the fovea. The scanswere centered on the fixation point by instructing thepatient to fixate the landmark point in the middle ofthe linear scan. Fixation was controlled by the videoimage of the central retina. Some patients had to beexcluded from the study because they were not able tomaintain stable fixation.In the horizontal scans, we measured retinal thick-

ness in the center of the fovea as well as 0.5 mm and1 mm from the center an both sides, corresponding toscan lines 14, 32, 50, 68, and 86. Two markers (whitearrowheads in Figure 1) were positioned manually onthe surface of the retina and the inner border of thehighly reflective outer band of pigment epithelium andchoriocapillaris. The distance between those two bor-ders consistently corresponds to total retinal thicknessin histologic comparison.25,26 Figure 1 gives an exam-ple of the measurement locations in a normal and adiabetic scan. The respective mean values of threerepeated horizontal scans were entered into furtheranalysis. The vertical scans were evaluated in a sim-ilar fashion, i.e., in the center as well as 0.5 and 1.0mm on either side. Thus, nine distinct thickness valuesfor each eye were available for analysis. In addition,average retinal thickness was calculated as the meanvalue of the nine measuring locations.

Angiography

Fluorescein angiograms of both eyes were obtainedsubsequent to injection of 5 mL of 20% sodium flu-orescein intravenously. Standard 30° fundus photo-graphs were taken, corresponding to fields 1, 2, 4, and5 of the ETDRS chart. Fluorescein leakage, capillarydestruction, and cystoid changes were assessed in the

venous phase of the angiogram. Nine subfields weredefined, corresponding to the ETDRS classification ofthe fluorescein angiograms27 (Figure 2). Leakage, cys-toid changes, and capillary nonperfusion were gradedin each subfield (Tables 1 and 2). In the central sub-field (subfield 1), the deformation of the contour of thefoveal avascular zone (FAZ) replaced the nonperfu-sion score. By summing the single score values, wecalculated a total score separately for leakage, cystoidspaces, and nonperfusion. The total score could rangefrom 0 to 36 for each of the three parameters. Theangiograms were graded by an experienced retinaspecialist, who was masked to the results of the OCTmeasurements.

Statistics

Only one randomly chosen eye of each patient wasincluded to avoid bias by dependent samples. If onlyone eye fulfilled the inclusion criteria, that eye waschosen. Group comparisons were done with the Wil-coxon–Mann–Whitney test for independent samples.Linear correlation was analyzed using the Pearsonproduct-moment correlation coefficient (r) and the

Fig. 1. Horizontal optical coherence tomography scan through thecenter of the fovea in a normal (A) and a diabetic (B) subject. The tipof the top white arrows marks the inner surface of the retina. The tip ofthe bottom white arrows marks the border of the outer high reflectiveband, corresponding to the surface of the retinal pigment epithelium.

761RETINAL THICKNESS IN DIABETIC RETINOPATHY • GOEBEL AND KRETZCHMAR-GROSS

matching significance of the correlation (P). The vari-ation coefficient (VAR) was calculated from the re-peated scans separately for each horizontal location asthe percentage of variance (standard deviation [SD])relative to the mean of retinal thickness at the singlelocation. The mean variation coefficient was com-puted as the mean of the single coefficients for allpatients. It provides a measure of intravisit reproduc-ibility. Visual acuity in Snellen equivalents was con-verted to logMAR units (logarithm of minimum angleof resolution) before being entered into analysis.Sensitivity and specificity of the OCT thickness

measurements were calculated using mean retinalthickness of our control population#2 SD as a cut-offvalue. The reference standard in our study was thepresence or absence of either leakage (total leakagescore) in fluorescein angiography or clinically signif-icant macular edema (CSME as defined by ETDRS) inslit-lamp biomicroscopy.

Results

Fourteen patients with diabetes were excluded ow-ing to improper fixation during OCT measurements. Atotal of 136 patients could be further evaluated. Thedemographics are given in Table 3. As expected,patients with type 1 diabetes were significantlyyounger. Duration of diabetes mellitus and the level ofHbA1c did not differ significantly between the dia-betic groups. There was no relationship between reti-nal thickness as measured with OCT and HBA1c orthe duration of diabetes mellitus. One hundred sixteenof 136 eyes (85%) showed CSME in slit-lampbiomicroscopy.Reproducibility of the OCT measurements was high

for diabetic and control subjects. The mean variationcoefficient of retinal thickness in the fovea was 2.3%(7 !m) in diabetic subjects and 3.3% (5 !m) in thecontrol group. One millimeter temporal and 1 mmnasal to the fovea, the variation coefficient was 2.4%

Fig. 2. Grid centered on the fovea and dividing the macula into ninesubfields. The radius of the innermost circle corresponds to one-third ofthe disk size, approximately 500 !m. The radius of the two outercircles measures 1 disk diameter and 2 disk diameters, respectively.

Table 1. Grading of Fluorescein Angiograms: AssigningScore Values

Score Grading

0 Absent, normal1 Questionable2 Present3 Moderate4 Severe

Table 2. Grading of Fluorescein Angiograms:Calculation of Total Leakage, Cystoid Edema, and

Nonperfusion Score

Score, 0–36 Field

LeakageExtent 1–9

Cystoid edemaCystoid spaces 1–9

NonperfusionDefects in contour of foveal avascular zone 1Nonperfusion areas 2–9

Table 3. Demographics of the Subject Groups

DemographicsNormal,n $ 30*

DiabetesTotal,

n $ 136†

DiabetesType 1,n $ 15

DiabetesType 2 NIDDM,

n $ 55

DiabetesType 2 IDDM,

n $ 66

Age, yr 53 ! 20 64 ! 12 42 ! 11‡ 66 ! 9 66 ! 8Diabetes, yr — 16 ! 8 20 ! 7 13 ! 9 17 ! 8HbA1c, % — 8.4 ! 1.7 8.0 ! 1.6 8.1 ! 1.7 8.6 ! 1.7

Values are expressed as mean ! SD.* 12 women; 18 men.† 71 women; 65 men.‡ p " 0.01.NIDDM, non–insulin-dependent diabetes mellitus; IDDM, insulin-dependent diabetes mellitus.

762 RETINA, THE JOURNAL OF RETINAL AND VITREOUS DISEASES ● 2002 ● VOLUME 22 ● NUMBER 6

(8 !m) and 1.4% (5 !m) in diabetic subjects and 2.4%(6 !m) and 2.2% (6 !m) in the control group.Mean retinal thickness in the fovea of 136 diabetic

patients was significantly greater than foveal thicknessof the control group, namely 307 ! 136 !m [mean !SD] as compared to 153 ! 15 !m (mean ! SD). Inthe extrafoveal locations, mean retinal thicknessranged from 337 !m to 353 !m in diabetic subjectsand 217 !m to 270 !m in the control group (Figure 3,A and B). The differences between diabetic and con-trol subjects were significant in every location (P "0.01), but were most pronounced in the fovea (P "0.001). Average retinal thickness (mean ! SD) was342 ! 97 !m in diabetic subjects and 225 ! 23 !min the control group (P " 0.01). Retinal thickeningin the subgroup of 116 diabetic eyes with CSMEwas even more pronounced, with thickness values of333! 131 !m in the fovea and 356! 92 !m average.However, foveal and average retinal thickness in thesubgroup of 20 diabetic eyes without CSME did notdiffer significantly from control eyes (159 ! 15 !mfoveal thickness; 233 ! 20 !m average thickness).We examined the correlation between retinal thick-ness in the center of the fovea and the eight extrafo-veal locations to determine if foveal thickening alone

may be an indicator of thickening in the entire macularregion. The correlation coefficient (r) between foveaand extrafoveal locations was high and ranged be-tween r $ 0.89 and r $ 0.95.In diabetic patients, correlation between foveal ret-

inal thickness and visual acuity in logMAR units wassmall (r $ 0.39, P " 0.00001; Figure 4A). However,when considering only eyes lacking capillary destruc-tion in the angiogram (total nonperfusion score $ 0),the correlation was better (r $ 0.51, P $ 0.002, n $34; Figure 4B). The correlation coefficient improvedto r $ 0.79 (P $ 0.0008) when further excluding eyeswith any degree of cataract from the subset, althoughthe remaining number was small (n $ 14; Figure 4C).Mean foveal thickness in the control group and dia-betic subjects with good vision (visual acuity%20/30)was significantly smaller than in the subgroups withdecreased visual acuity (Figure 5).Fluorescein angiograms could be evaluated in 112

of the 136 eyes chosen for analysis. In 14 eyes, thequality of the angiograms was not adequate for ourgrading scheme. In 10 patients, angiographic datawere not available, mainly because patients refusedinvasive procedures or because internal diseases pre-cluded angiography. The mean leakage score was 14.8(range, 0–33). Only four eyes had no leakage in theangiogram (leakage score $ 0). The mean score ofcystoid changes was 4.2 (range, 0–29). Cystoidchanges were present in 50% of the eyes. Cystoidspaces were more prominent in the central fields1–5compared to the outer fields.6–9 Only 16 eyes showedany cystoid changes in the outer fields, usually incombination with pronounced macrocystoid spaces inthe central fields. The mean nonperfusion score was3.3 (range, 0–19). Thirty-four eyes (30%) had nononperfusion areas or defects in the contour of thefoveal avascular zone (nonperfusion score $ 0).There was an intermediate relationship between av-

erage retinal thickness as measured by OCT and theleakage score derived from the angiograms (Figure6A). The correlation coefficient (r $ 0.44) was sig-nificant (P " 0.00001). The correlation between av-erage retinal thickness and the extent of cystoidchanges in the angiograms was slightly smaller (r $0.40, P $ 0.00002; Figure 6B). As expected, therewas no correlation between retinal thickness and non-perfusion areas (nonperfusion score) in the angio-grams (r $ 0.004, P $ 0.96).We calculated the sensitivity and specificity of de-

tecting any degree of fluorescein leakage from mea-suring foveal retinal thickness with OCT, regardingthickness values above 183 !m (mean of controls #2*SD) as pathologic (Table 4). Sensitivity was 81%and specificity was 94%. When using average retinal

Fig. 3. A, Horizontal cross-section of retinal thickness in diabeticsubjects (n $ 136; straight line) and control subjects (n $ 30; dottedline). Smoothed curve of mean values at five different locations:d(temporal)$ 1.0 mm, d(temporal)$ 0.5 mm, center of the fovea,d(nasal)$ 0.5 mm, and d(nasal)$ 1.0 mmB. B, Vertical cross-sectionof retinal thickness in diabetic and control subjects.

763RETINAL THICKNESS IN DIABETIC RETINOPATHY • GOEBEL AND KRETZCHMAR-GROSS

thickness as a parameter, regarding values above 271!m (mean of controls # 2*SD) as pathologic, sensi-tivity was 73% and specificity 100%. Taking CSMEas the reference standard, sensitivity was 89% andspecificity 96% for foveal retinal thickness and 80%(sensitivity) and 100% (specificity) for average retinalthickness.

Discussion

In the current study of 136 patients with diabeticmaculopathy, we analyzed the relationship betweenfluorescein angiography, CSME detected by slit-lampbiomicroscopy, visual function, and retinal thickeningas measured by OCT. Macular edema is defined as anincrease in retinal volume. Because of the anatomicarchitecture, this increase in volume is represented byan increase in retinal thickness.7 OCT is currently themost precise technique for the measurement of retinalthickness in vivo. In our study, we could confirm theprecision of OCT measurements in a clinical settingnot only in controls but also in a large set of diabeticpatients. The mean variation coefficient in differentmeasurement locations varied between 2.2 and 3.3%in the control group and 1.4 and 2.4% in diabeticsubjects. Those values compare well with intrasessionreproducibility figures of 1.2%,28 3.2% to 8.1%,29 and7.2%20 in control subjects in the literature. We did notfind greater measurement variability in our sample ofdiabetic subjects compared to the control group. Ourreproducibility figures in the diabetic group were evenbetter than values of 8.1% to 8.9%20 reported previ-ously. This is particularly significant because our di-abetic patients showed more pronounced retinal thick-ening as compared with other studies.19,20 We

Fig. 4. Visual acuity in logMar units plotted against foveal retinalthickness. A, Total diabetic group (n $ 136) (correlation: r $ 0.39). B,No destruction of foveal avascular zone (FAZ) subgroup (n $ 34)(correlation: r $ 0.54). C, No destruction of FAZ and no cataractsubgroup (n $ 14) (correlation: r $ 0.79). Regression line and 95%confidence interval are shown.

Fig. 5. Foveal retinal thickness in diabetic subjects stratified by visualacuity compared to the control group. Mean (box) and standard devi-ation (whiskers) are shown. Asterisks indicate significance with refer-ence to the control group. **P " 0.01; ***P " 0.001.

764 RETINA, THE JOURNAL OF RETINAL AND VITREOUS DISEASES ● 2002 ● VOLUME 22 ● NUMBER 6

speculate that the use of the prototype instrument witha scan time of 2.5 seconds in previous studies mayhave contributed to the greater variability. In ourstudy, the required fixation time of 1 second with thecommercial instrument and latent small changes inscan placement due to eye movements did not bear amajor influence on the precision of OCT measure-ments. However, we carefully controlled fixation anddid exclude 14 of 150 diabetic patients (9%) in theOCT session after enrollment because of fixationproblems.Foveal retinal thickness of our control population

(153 ! 15 !m) was consistent with values of 147 !17 !m,19 152 ! 21 !m,20 154 ! 13 !m,29 152 ! 17!m,30 and 142 ! 18 !m31 previously reported by usand by other authors using the same technique. In ourstudy, retinal thickness as determined by OCT wasrepresented by nine discrete values for each eye. Al-though we may have missed localized thickening out-side the nine measuring points, the high correlation

between foveal and average thickness as well as reti-nal thickness in the eight extrafoveal locations (r %0.89) makes it unlikely that relevant thickening in themacula would leave foveal thickness unchanged. Fur-thermore, this correlation indicates that measurementsof thickness in the fovea alone might be sufficientwhen screening for CSME. Our results thus corrobo-rate findings by Hee and coworkers,20 who also estab-lished a high correlation between central macular andaverage foveal thickness (r $ 0.98). We analyzed ourscans manually because at the time the study wasdesigned, a software version calculating and exportingan edema map20 was not available and the built-inalgorithm to evaluate retinal thickness proved unreli-able under certain conditions. In macular edema aswell as in the central fovea of normal subjects, thereflectivity of the inner parts of the retina is close tothe noise level, so the software algorithm for thicknessanalysis may fail in some positions of the scan.29Furthermore, intraretinal exudates and hemorrhagesmay produce reflections higher than the pigment epi-thelial layer. Under those conditions, the softwarealgorithm cannot reliably distinguish between highreflectivity from within the retina and high reflectivityfrom the pigment epithelium. In manual measure-ments, consistency can be verified easily by interpo-lating from neighboring scan positions. Improvementsof the algorithm may lead to better performance indifficult situations, but still, some degree of manualcontrol should be implemented.In our study, lower levels of visual acuity ("20/30

Snellen equivalent) were significantly associated witha marked increase in retinal thickness. However, theoverall inverse correlation (correlation coefficient r $0.39) we found between central retinal thickening inOCT and visual acuity was considerably worse than in

Fig. 6. A, Total leakage score plotted against mean retinal thickness (n$ 112; correlation: r $ 0.44). Regression line (solid line) and 95% confidenceinterval (dotted line) are shown. B, Total cystoid edema score plotted against mean retinal thickness (n $ 112; correlation: r $ 0.40). Regression line(solid line) and 95% confidence interval (dotted line) are shown.

Table 4. Eyes With/Without Leakage on Angiography orClinically Significant Macular Edema (CSME)

Retinal Thicknessby OCT, !m

LeakageAngiogram

NormalAngiogram

CSMEClinical

NormalClinical

Fovea%183 88 2 103 2"183 20 32 13 48

Mean%271 79 0 93 0"271 29 34 23 50

Values are expressed as no. of eyes.Number of eyes with/without leakage or CSME was used for

calculation of sensitivity and specificity. Normal groups includecontrols (controls are assumed to have no leakage in angiogra-phy).

OCT, optical coherence tomography.

765RETINAL THICKNESS IN DIABETIC RETINOPATHY • GOEBEL AND KRETZCHMAR-GROSS

previous reports stating correlation coefficients of r $0.67,19 r $ 0.89,20 and r $ 0.61.32 However, thecorrelation improved and reached values comparableto previous studies when excluding from analysis pa-tients with angiographically visible ischemia (r $0.51) and cataract (r $ 0.79). This observation con-firms that visual acuity in diabetic patients not onlydepends on edema formation but also on capillarydestruction in the macula. Similarly, in a study byAntcliff and coworkers33 the correlation between vi-sual acuity and retinal thickness measured by OCTwas only intermediate (r$ 0.53), probably because theirsample of patients with cystoid macular edema fromuveitis also included subjects with media opacities andvarious degrees of ischemia from occlusive vasculitis. Inconclusion, visual acuity seems to be a poor predictor ofretinal thickening and, hence, macular edema in ad-vanced stages of diabetic maculopathy.Slit-lamp biomicroscopy in combination with fluo-

rescein angiography is the standard technique for thedetection, analysis, and follow-up of diabetic macularedema. In our study, increased retinal thickness wasmore closely linked with CSME than with angio-graphically detectable leakage. Although areas ofleakage in fluorescein angiography indicate localizedblood–retinal barrier breakdown and thus indicatesites where macular edema is likely to occur, fluores-cein angiography does not provide a measure of reti-nal thickening itself.33 Grading the amount of edemain fluorescein angiograms is unreliable because edemaformation is a variable process that depends on thebalance between the rate of leakage and the amount ofreabsorption of fluid. Correspondingly, a correlationbetween the degree of retinal thickening and theamount of leakage or cystoid changes was present butnot very close in our data. As far as cystoid changesare concerned, the correlation is further negativelyinfluenced by the distribution of the cystoid edemascore being considerably skewed toward lower values.We found the sensitivity and specificity of detecting

CSME by measuring retinal thickness in the fovea tobe very good (89% sensitivity and 96% specificity).Foveal thickness alone was a better indicator ofCSME than average retinal thickness as far as sensi-tivity is concerned. This is due to the predominance offoveal over more peripheral macular edema, a factfurther corroborated by the greater difference in meanretinal thickness between diabetics and control in thecentral fovea compared to the extrafoveal measuringlocations. So far, sensitivity data regarding OCT mea-surements in diabetic retinopathy are largely missing.Although Hee and coworkers20 did not explicitly statesensitivity, a calculation based on their data renders asensitivity of 74% for the detection of CSME by

measuring average foveal thickness with OCT. Sensi-tivity values reported for the volumetric analysis ofmacular edema with the Heidelberg Retina Tomo-graph vary between 58%34 and 79%,11 but the repro-ducibility is significantly worse than in OCT measure-ments. For the RTA, sensitivity values between 69%10

and 100%15 as well as specificity values of 88% and69% have been calculated. Although our data seem tobe very favorable in comparison, the results may in-clude some bias owing to the study design. The clin-ical setting overemphasizes diabetic subjects withmacular edema compared to the general population.Eighty-five percent of eyes showed CSME on slit-lamp biomicroscopy. Thus, the number of false-posi-tive OCT measurements may have increased and spec-ificity may have deteriorated if more eyes with lesssevere maculopathy had been included, as some de-gree of retinal thickening may be present even in eyeswithout CSME.30In summary, we were able to demonstrate that OCT

shows excellent reproducibility not only in controlsbut also in diabetic patients and has considerablepotential for diagnosing sight-threatening macularedema by measuring retinal thickness. Compared withangiography, OCT allows us to quantify retinal thick-ness and, hence, macular edema in diabetic retinopa-thy objectively and noninvasively. Therefore, OCTmay become a valuable additional tool not only in thefollow-up of macular edema but also in screeningprograms for diabetic retinopathy.Key words: diabetic retinopathy, macular edema,

optical coherence tomography.

References1. Klein R, Klein BE, Moss SE. Visual impairment in diabetes.

Ophthalmology 1984;91:1–9.2. Smith RT, Lee CM, Charles HC, Farber M, Cunha-Vaz JG.

Quantification of diabetic macular edema. Arch Ophthalmol1987;105:218–222.

3. Nussenblatt RB, Kaufman SC, Palestine AG, et al. Macularthickening and visual acuity. Measurement in patients withcystoid macular edema. Ophthalmology 1987;94:1134–1139.

4. Early Treatment Diabetic Retinopathy Study ResearchGroup. ETDRS report number 10. Grading diabetic retinop-athy from stereoscopic color fundus photographs—an exten-sion of the Modified Airlie House classification. Ophthalmol-ogy 1991;98:786–806.

5. Early Treatment Diabetic Retinopathy Study ResearchGroup. Fluorescein angiographic risk factors for progressionof diabetic retinopathy. ETDRS report number 13. Ophthal-mology 1991;98:834–840.

6. Varano M, Scassa C, Ripandelli G, Capaldo N. New diag-nostic tools for macular edema. Doc Ophthalmologica 1999;97:169–175.

7. Lobo C, Bernardes R, Faria de Abreu JR, Cunha-Vaz JG.Novel imaging techniques for diabetic macular edema. DocOphthalmol 1999;97:341–347.

766 RETINA, THE JOURNAL OF RETINAL AND VITREOUS DISEASES ● 2002 ● VOLUME 22 ● NUMBER 6

8. Lobo C, Bernardes R, Santos F, Cunha-Vaz JG. Mappingretinal fluorescein leakage with confocal scanning laser flu-orometry of the human vitreous. Arch Ophthalmol 1999;117:631–637.

9. Lobo C, Bernardes R, Cunha-Vaz JG. Alterations of theblood–retinal barrier and retinal thickness in preclinical ret-inopathy in subjects with type 2 diabetes. Arch Ophthalmol2000;118:1364–1369.

10. Shahidi M, Ogura Y, Blair NP, Rusin MM, Zeimer R. Retinalthickness analysis for quantitative assessment of diabeticmacular edema. Arch Ophthalmol 1991;109:1115–1119.

11. Zambarakji HJ, Amouku WM, Vernon SA. Volumetric anal-ysis of early macular edema with the Heidelberg RetinaTomograph in diabetic retinopathy. Ophthalmology 1998;105:1051–1059.

12. Hudson Ch, Flanagan JG, Turner GS, McLeod D. Scanninglaser tomography Z profile signal width as an objective indexof macular retinal thickening. Br J Ophthalmol 1998;82:121–130.

13. Zambarakji HJ, Vernon SA, Spencer AF, Amoaku WMK.Reproducibility of volumetric measurements in diabetic pa-tients with the Heidelberg Retina Tomograph. Doc Ophthal-mol 1999;97:145–156.

14. Zeimer R, Shahidi M, Mori M, et al. A new method for rapidmapping of the retinal thickness at the posterior pole. InvestOphthalmol Vis Sci 1996;39:1994–2001.

15. Yasukawa T, Kiryu J, Tsujikawa A, et al. Quantitative anal-ysis of foveal retinal thickness in diabetic retinopathy withthe scanning retinal thickness analyser. Retina 1998;18:150–155.

16. Weinberger D, Axer-Siegel R, Landau D, Yassur Y. Retinalthickness variation in the diabetic patient measured by theretinal thickness analyzer. Br J Ophthalmol 1998;82:1003–1006.

17. Oshima Y, Emi K, Yamanishi S, Motokura M. Quantitativeassessment of macular thickness in normal subjects and pa-tients with diabetic retinopathy by scanning retinal thicknessanalyser. Br J Ophthalmol 1999;83:54–61.

18. Puliafito CA, Hee MR, Lin CP, et al. Imaging of maculardiseases with optical coherence tomography. Ophthalmology1995;102:217–229.

19. Hee MR, Puliafito CA, Wong C, et al. Quantitative assess-ment of macular edema with optical coherence tomography.Arch Ophthalmol 1995;113:1019–1029.

20. Hee MR, Puliafito CA, Duker JS, et al. Topography ofdiabetic macular edema with optical coherence tomography.Ophthalmology 1998;105:360–370.

21. Otani T, Kishi Sh, Maruyama Y. Patterns of diabetic macularedema with optical coherence tomography. Am J Ophthalmol1999;127:688–693.

22. Early Treatment in Diabetic Retinopathy Study ResearchGroup. Photocoagulation for diabetic macular edema. EarlyTreatment Diabetic Retinopathy Study Report Number 1.Arch Ophthalmol 1985;103:1796–1806.

23. Swanson BA, Izatt JA, Hee MR, et al. In vivo retinal imagingby optical coherence tomography. Opt Lett 1993;18:1864—1866.

24. Hee MR, Izatt JA, Swanson EA, et al. Optical coherencetomography of the human retina. Arch Ophthalmol 1995;113:325–332.

25. Toth CA, Narayan DG, Boppart SA, et al. A comparison ofretinal morphology viewed by optical coherence tomographyand by light microscopy. Arch Ophthalmol 1997;115:1425–1428.

26. Chauhan DS, Marshall J. The interpretation of optical coher-ence tomography images of the retina. Invest Ophthalmol VisSci 1999;40:2332–2342.

27. Early Treatment Diabetic Retinopathy Study ResearchGroup. ETDRS report number 11. Classification of diabeticretinopathy from fluorescein angiograms. Ophthalmology1991;98:807–822.

28. Koozekanani D, Toberts C, Katz SE, Herderick EE. Interses-sion repeatability of macular thickness measurements withthe Humphrey 2000 OCT. Invest Ophthalmol Vis Sci 2000;41:1486–1491.

29. Baumann M, Gentile RC, Liebmann JM, Ritch R. Reproduc-ibility of retinal thickness measurements in normal eyes usingoptical coherence tomography. Ophthalmic Surg Lasers1998;29:280–285.

30. Schaudig UH, Glaefke C, Scholz F, Richard G. Opticalcoherence tomography for retinal thickness measurement indiabetic patients without clinically significant macularedema. Ophthalmic Surg Lasers 2000;31:182–186.

31. Göbel W, Hartmann F, Haigis W. The correlation betweenretinal thickness and axial length as well as age, measuredby optical coherence tomography. Ophthalmologe 2001;98:157–162.

32. Tomohiro O, Kishi S, Maruyama Y. Pattern of diabeticmacular edema with optical coherence tomography. Am JOphthalmol 1999;127:688–693.

33. Antcliff RJ, Stanford MR, Chauhan DS, et al. Comparisonbetween optical coherence tomography and fundus fluo-rescein angiography for the detection of cystoid macularedema in patients with uveitis. Ophthalmology 2000;107:593–599.

34. Zambarakji HJ, Butler TKH, Vernon SA. Assessment of theHeidelberg Retina Tomograph in the detection of sight-threatening diabetic maculopathy. Eye 1999;13:136–144.

767RETINAL THICKNESS IN DIABETIC RETINOPATHY • GOEBEL AND KRETZCHMAR-GROSS