Since the introduction of optical coherence tomography

Transcription

1 Multidisciplinary Ophthalmic Imaging Repeatability of Peripapillary Retinal Nerve Fiber Layer and Inner Retinal Thickness Among Two Spectral Domain Optical Coherence Tomography Devices Juliane Matlach, 1 Martin Wagner, 2,3 Uwe Malzahn, 2 and Winfried Göbel 1 1 Department of Ophthalmology, University of Würzburg, Germany 2 Institute of Clinical Epidemiology and Biometry, University of Würzburg, Germany 3 Comprehensive Heart Failure Center, University of Würzburg, Germany Correspondence: Juliane Matlach, Department of Ophthalmology, Josef-Schneider-Str. 11, Würzburg, Germany; j.matlach@augenklinik. uni-wuerzburg.de. Submitted: June 20, 2014 Accepted: September 3, 2014 Citation: Matlach J, Wagner M, Malzahn U, Göbel W. Repeatability of peripapillary retinal nerve fiber layer and inner retinal thickness among two spectral domain optical coherence tomography devices. Invest Ophthalmol Vis Sci. 2014;55: DOI: /iovs PURPOSE. To compare measurement of macular inner retina and peripapillary retinal nerve fiber layer (prnfl) thickness using two spectral-domain optical coherence tomography (SD- OCT) devices in glaucoma patients, patients with ocular hypertension, idiopathic and atypical Parkinson disease, and healthy controls. METHODS. A total of 171 eyes of 146 participants underwent successful prnfl and macular scanning and automated measurement of ganglion cell layer þ inner plexiform layer (GCL- IPL) using Cirrus HD-OCT or retinal nerve fiber layer þ GCL-IPL (RNFL GCL-IPL) using RTVue-100. Macular RNFL was added to the GCL-IPL thickness measured by Cirrus and was compared to the RNFL GCL-IPL thickness of the RTVue in the corresponding Cirrus sectors. Intraclass correlation coefficient (ICC) was calculated to determine repeatability of three consecutive measurements; ICC and Bland-Altman analysis to assess agreement between OCTs; Pearson s correlation coefficient to assess strength of linear correlation. RESULTS. Repeatability of average macular RNFL GCL-IPL thickness measurement was excellent with an ICC of for Cirrus and for RTVue. Repeatability was also good for average prnfl thickness measurements. Both instruments demonstrated a good consistency in measurements with ICC values ranging from to for macular RNFL GCL-IPL and to for prnfl thickness. CONCLUSIONS. Measurement of prnfl and macular RNFL GCL-IPL thickness has a high degree of repeatability for both OCT devices. Despite a high correlation between measurements of the two OCT devices and fair to excellent ICC values representing a high consistency in the measurements of the two devices, RTVue measured a thicker macular RNFL GCL-IPL and prnfl thickness. Keywords: OCT, ganglion cell layer, inner retina, RNFL, Cirrus, RTVue, repeatability, agreement Since the introduction of optical coherence tomography (OCT) in 1991, 1 it has revolutionized imaging in ophthalmology, aiding clinicians and researchers in understanding and visualizing retinal and choroidal pathologies as well as providing objective measurements of peripapillary retinal nerve fiber layer (prnfl) and macular thickness. 2,3 More recently, with the further development of spectral-domain (SD)-OCT, acquisition speed and image resolution have greatly increased; thus, allowing for precise segmentation and measurement of intraretinal layers, particularly the ganglion cell (GCL) and inner plexiform layers (IPL). Nearly in vivo, high-resolution and noninvasive imaging has revolutionized the clinicians point of view by providing images of the posterior pole and measurement of the prnfl or macular thickness. Different commercially available OCT devices are used for diagnosing and monitoring glaucoma 4 6 and retinal diseases. In addition, OCT is now widely accepted in neurology 7 for imaging of optic neuropathy in multiple sclerosis 8,9 and is used in clinical studies for degenerative disorders of the central nervous system such as Parkinson 10 and Alzheimer disease. 11 Unfortunately, different SD-OCT devices use manufacturerspecific segmentation algorithms leading to widely variable prnfl and macular thickness measurements and false interpretation of disease improvement or progression. 3,9,10,12 25 Intradevice repeatability and interdevice agreement must be excellent in order to have reliable data especially when new segmentation algorithms such as ganglion cell analysis GCL/IPL are investigated in different studies using several OCT devices. Repeatability and reproducibility of prnfl thickness, 9,13,14,16,17 overall and central macular thickness 12,15,18,22 in healthy participants, glaucoma suspects, and glaucoma patients as well as patients with retinal and neurodegenerative diseases have been investigated using different SD-OCT devices or in comparison with time domain (TD)-OCT. So far, only few studies investigated repeatability of macular inner retinal thickness (GCL/IPL or ganglion cell complex) measurements using SD-OCT To our knowledge, there is no study comparing macular inner retinal thickness using different OCT instruments and modifying the segmentation sectors that has comparable results. Copyright 2014 The Association for Research in Vision and Ophthalmology, Inc. j ISSN:

2 RNFL and Inner Retinal Thickness Using Two SD-OCTs IOVS j October 2014 j Vol. 55 j No. 10 j 6537 Given the value of variable thickness measurement among different SD-OCT devices, the purpose of this study was to compare automated measurement of macular inner retina and prnfl thickness using two different SD-OCT devices (Cirrus HD-OCT; Carl Zeiss Meditec, Inc., Dublin, CA, USA; and RTVue; Optovue, Inc., Fremont, CA, USA); to assess repeatability of measurements; and to evaluate agreement between both OCT devices in preperimetric and perimetric glaucoma patients, patients with ocular hypertension (OHT), patients with idiopathic and atypical Parkinson disease (PD), and normal controls. We focused on intradevice repeatability of scans and interdevice agreement between Cirrus and RTVue rather than on thickness differences between different disease groups and healthy normals. A wide range of diseases has been included to have representative thickness values and a large number of cases. PATIENTS AND METHODS Participants The study protocol was approved by the local ethics committee of the University of Würzburg, Germany (reference 116/12, 138/12). The research followed the tenets of the Declaration of Helsinki. Written informed consent was obtained from all participants after explanation of the nature and possible consequences of the study. All participants were included between August 2012 and July 2013 at the Department of Ophthalmology, University of Würzburg, Germany. In total, 165 enrollees were examined of which 19 had to be excluded due to segmentation errors, severe media opacities, and poor scan quality. Thus, OCT data of 171 eyes of 146 participants were suitable for analyses. To provide a study sample to assess the peripapillary RNFL and inner retinal thickness over a sufficiently wide spectrum, patients with different disorders were included. Of these, 51 eyes of 43 healthy volunteers, 63 eyes of 63 patients with perimetric glaucoma, three eyes of three patients with preperimetric glaucoma, two eyes of two patients with ocular hypertension, 46 eyes of 30 patients with idiopathic PD, and six eyes of five patients with atypical PD were included. Inclusion criteria were male or female participants aged 18 years or older with a spherical equivalent between 8.00 and þ6.50 diopters (D). Both eyes were scanned if inclusion and exclusion criteria were fulfilled. Exclusion criteria were bestcorrected visual acuity (BCVA) < 20/40, evidence of macular pathologies including AMD, macular hole, macular pucker or vitreomacular traction syndrome, vascular or inflammatory diseases, hereditary retinal dystrophies and diabetic or hypertensive retinopathy with macular changes, optic nerve neuropathy other than glaucomatous optic neuropathy, previous surgery for macular or vitreoretinal diseases, previous ocular surgeries or laser treatment within 4 months prior examination such as cataract extraction, refractive surgery, retinal photocoagulation or glaucoma surgery (glaucoma patients only), significant media opacities, poor quality scans with signal strength < 6 (Cirrus) and < 45 (RTVue), segmentation errors, and eye motion artifacts. Typical glaucomatous optic nerve signs include glaucomatous optic nerve cupping, flame-shaped or splinter hemorrhages at the optic disc margin, nasalization, and bayoneting of the blood vessels, pallor of the neuroretinal rim, and retinal nerve fiber layer defects. Perimetric patients, defined as having characteristic glaucomatous optic nerve head changes and visual field defects, were distinguished from preperimetric patients without visual field defects and patients with ocular hypertension without any glaucomatous changes of the optic nerve or visual field defects despite an intraocular pressure (IOP) outside the normal range. All participants were asked for medical and ocular history and completed a standard ophthalmic examination including autorefraction, visual acuity testing, slit-lamp biomicroscopy, and indirect ophthalmoscopy for fundus examination as well as Goldmann applanation tonometry for IOP measurement. Glaucoma patients additionally underwent visual field testing using commercial devices (Octopus 900 or 101 threshold program G1 Haag Streit, Koeniz, Switzerland) conducted within 3 months prior examination. Unreliable visual fields with false-positive rates of >30% were repeated or discarded if not possible. Optical Coherence Tomography Imaging All OCT scans for both devices were performed by a single trained operator (JM) and were taken three times consecutively on the same day to study repeatability of measurements. Images with poor quality, segmentation errors, or artifacts were excluded. Cirrus HD-OCT 4000 (software version 6.02) and RTVue- 100 (software version 6.3) were used to assess prnfl and macular inner retina thickness. For the Cirrus OCT, the optic disc cube and the macular cube scan protocols consisting of 200 vertical and 200 horizontal scan points within a cube measuring mm were used and processed with the optic nerve head (ONH) and RNFL analysis or ganglion cell analysis algorithm, respectively. A circumpapillary scan of 3.46 mm diameter around the center of the optic disc was generated to measure prnfl thickness. The thickness of the macular GCL was automatically calculated in predefined sectors. The segmentation of the inner retinal layers as identified by the GCA algorithm automatically summarizes the ganglion cell (GCL) and the inner plexiform layer (IPL) based on the three-dimensional (3D) data from the macular cube protocol. The difference between the outer boundary of the macular RNFL and IPL yields the segmented GCL-IPL (Fig. 1B). The average, minimum, and six sectors superior, superonasal, superotemporal, inferior, inferonasal and inferotemporal defined for the thickness map of the GCA are measured within an elliptical annulus with exclusion of the fovea (Fig. 1A). The dimension of the elliptical annulus encompasses a mm inner radius centered on the fovea and a mm outer radius conforming closely to the anatomic arrangement of the ganglion cells. For the RTVue, macular OCT images were obtained using the EMM5 protocol for inner retina thickness measurement. The algorithm segments the macular RNFL, GCL, and IPL, and summarizes it to form the ganglion cell complex (GCC) in the nine Early Treatment Diabetic Retinopathy Study (EDTRS) sectors around the fovea using a composite of þ A-scans within a mm scan area (Figs. 1C, 1D). Peripapillary RNFL thickness and optic nerve head parameters were determined using the ONH and 3D disc protocol. Average, superior, nasal, inferior, temporal and clock-hour prnfl thicknesses are calculated using a 3.45-mm diameter circle of 4.9 mm around the optic disc after defining the optic disc margin on the basis of the 3D or video image of the optic disc. The disc margin was redrawn and new anchoring points were determined by the same examiner taking all OCT images. The analysis methods described above are device-specific standards for both instruments. Acquisition and segmentation protocols differ between Cirrus and RTVue and are shown in Table 1. Particularly, measurements of the inner retina are not directly comparable between RTVue and Cirrus. The ganglion cell complex or RNFL GCL-IPL thickness as generated by the RTVue is measured from the internal limiting membrane to the

3 RNFL and Inner Retinal Thickness Using Two SD-OCTs IOVS j October 2014 j Vol. 55 j No. 10 j 6538 FIGURE 1. Imaging with SD-OCT of a healthy control. Cirrus OCT: False-color retinal thickness map displays averages of thickness measurement of GCL þ IPL in six sectors with an inner radius of mm and an outer radius of mm (A). Cirrus OCT: Segmentation of inner boundary of GCL (purple line) and outer boundary of IPL (yellow line) (B).

4 RNFL and Inner Retinal Thickness Using Two SD-OCTs IOVS j October 2014 j Vol. 55 j No. 10 j 6539 FIGURE 1. (continued) RTVue OCT: False-color retinal thickness map shows averages of thickness measurement for each of the 9 ETDRS sectors. Inner ring between 1 and 3 mm and outer ring between 3 and 5-mm diameter are segmented into 4 quadrants: superior outer/inner, inferior outer/ inner, temporal outer/inner, and nasal outer/inner (C). RTVue OCT: Segmentation of inner boundary of RNFL and outer boundary of IPL (D). outer boundary of the IPL in the nine ETDRS sectors; whereas the GCA or GCL-IPL thickness of the Cirrus OCT only summarizes the GCL and IPL within an elliptical annulus. To have comparable inner retina thickness measurements, raw processed data of macular RNFL GCL-IPL as segmented by the RTVue were transformed into the elliptical annulus used for the GCA of the Cirrus, thereby using the same dimensions of the inner and outer radius around the fovea (Fig. 2A). Then, automatically segmented macular RNFL measurements within the six sectors of the elliptical annulus were added to the GCL- IPL to have thickness values comparable with the RNFL GCL- IPL complex of the RTVue (Fig. 2B). In detail, raw processed data for inner retinal and total retinal thickness was exported from the RTVue. The resolution of the data matrix for inner and total retinal thickness was lm ( ¼ 644,809 data points in x- and y- axis in a mm area). The center of the fovea was identified as the minimum total retinal thickness in order to shift and align the data matrix. Hence, further analysis for inner retinal thickness was centered on the individual position of the fovea in the respective scans. Templates of the six GCL sectors which had the exact same dimensions as the respective GCL sectors in the Cirrus OCT were used to overlay the data matrix for inner retinal thickness. For each GCL sector, inner retinal thickness values within the template were electronically averaged, thus providing mean retinal thickness in individual GCL sectors comparable with the Cirrus OCT. Statistical Analysis Statistical analyses were performed using statistical software (SPSS for Windows 21.0; IBM Corporation, Armonk, NY, USA).

5 RNFL and Inner Retinal Thickness Using Two SD-OCTs IOVS j October 2014 j Vol. 55 j No. 10 j 6540 TABLE 1. Acquisition and Segmentation Protocols for Cirrus and RTVue Cirrus Software Version 6.02 RTVue Software Version 6.3 Macular scan Program Macular cube EMM5 Number of A-Scans (rectangle ) (composite þ ) Scan area Rectangle mm Rectangle 6 3 6mm Acquisition time 1.5 s 1.3 s Segmentation protocol Ganglion cell analysis: GCL þ IPL EMM5 with inner retina: RNFL þ GCL þ IPL Optic disc head and peripapillary scan Program Optic disc cube ONH Number of A-scans (rectangle ) (composite linear/circle) Scan area Rectangle mm Circle 4.9 mm Acquisition time 1.5 s 0.55 s Circumpapillary scan 3.46 mm (256 A-Scans) 3.45 mm (775 A-scans) GCL, ganglion cell layer; IPL, inner plexiform layer; RNFL, retinal nerve fiber layer. Mean prnfl and macular RNFL-GCL-IPL thickness of three scans taken consecutively by one single operator were calculated for both devices. Intraclass correlation coefficient (ICC) was used for assessment of repeatability of average prnfl and RNFL-GCL-IPL thickness measurements as well as of each quadrant or segments for both OCTs. Values of ICC based on a two-way mixed effects ANOVA model with a random patient factor and the OCT device as the fixed factor were calculated to assess the interrater reliability between both OCT devices. Furthermore, ICC values within a TABLE 2. Demographic Data of All Participants Eyes/participants, n 171/146 Healthy 51/43 Perimetric glaucoma 63/63 Preperimetric glaucoma 3/3 Ocular hypertension 2/2 Idiopathic Parkinson disease 46/30 Atypical Parkinson disease 6/5 Age, y Mean 6 SD (range) (40 78) Sex Male 75 (43.9) Female 96 (56.1) Eye Right 89 (52.0) Left 82 (48.0) BCVA, Snellen 20/16 23 (13.5) 20/ (61.4) 20/25 33 (19.3) 20/32 9 (5.3) 20/40 1 (0.6) Refractive error, D Emmetropia or D 125 (73.1) Myopia Moderate (< 2.00 to 6.00 D) 26 (15.2) Severe (< 6.00 D) 3 (1.8) Hyperopia Moderate (>þ2.00 to þ4.00 D) 15 (8.8) Severe (>þ4.00 D) 2 (1.2) Data are absolute values (%), mean 6 standard deviation, median (interquartile range [IQR]). one-way ANOVA model with a random patient factor were calculated to assess the intraobserver repeatability (with the OCT method as observer). We therefore proposed equal mean values for repeated measurements and used a splitting up of the total variance in within- and between patient variance. Alternatively, only in the sense of a sensitivity analysis, we calculated these ICC values within a two-way random effects ANOVA again with a random patient factor and accounting for three consecutive measurements at the same day, but not at fixed time points. Agreement between both instruments was assessed by a graphical approach, the Bland-Altman plot. At this for each participant the difference between the measurements for the two compared methods is calculated and plotted against the mean of the two measurements. Furthermore, calculated bias and limits of agreement are plotted as horizontal lines. Linear correlation between measurement variables was assessed by Pearson s product moment correlation coefficient. A paired t- test on the mean difference between the measurements was obtained on the two methods to verify the presence of a systematic bias. Values of P < 0.05 were considered statistically significant. RESULTS Demographics of study participants, BCVA, and refractive error are displayed in Table 2. Fifty-one eyes of 43 healthy controls, 71 eyes of 71 perimetric, preperimetric glaucoma patients, and patients with ocular hypertension as well as 52 eyes of 36 patients with idiopathic or atypical Parkinson disease were included. The majority of patients (74.9%) had a visual acuity of 20/20 or better and were emmetropic (up to D; 73.1%). Repeatability of Thickness Measurements Intraobserver repeatability was excellent for average and quadrant prnfl thickness measurements (Table 3) with an ICC averageprnfl of (95% confidence interval [CI]: ) for Cirrus and (95% CI: ) for RTVue (ICC > 0, P < ). Repeatability of macular RNFL GCL-IPL complex was also excellent for average thickness as well as for segments for both OCT devices. Intraclass correlation coefficient was (95% CI: ) and (95% CI: ) regarding average macular RNFL GCL-IPL complex for Cirrus and RTVue, respectively (Table 4; ICC > 0, P < ). Values of ICC are given with adequate precision (three digits), of course in any case the point estimate is located between the confidence bounds not coinciding with these.

6 RNFL and Inner Retinal Thickness Using Two SD-OCTs IOVS j October 2014 j Vol. 55 j No. 10 j 6541 TABLE 3. Instrument TABLE 4. Repeatability of Macular RNFL GCL-IPL Thickness for Cirrus OCT and RTVue Instrument Repeatability of prnfl Thickness for Cirrus OCT and RTVue ICC (95% CI) Cirrus n ¼ 162/171 Average ( ) Superior ( ) Temporal ( ) Nasal ( ) Inferior ( ) RTVue n ¼ 164/171 Average ( ) Superior ( ) Temporal ( ) Nasal ( ) Inferior ( ) Repeatability defined as three consecutive scans taken by one operator. ICC (95% CI) Cirrus n ¼ 165/171 Average ( ) Superior ( ) Superonasal ( ) Superotemporal ( ) Inferior ( ) Inferonasal ( ) Inferotemporal ( ) RTVue n ¼ 141/171 Average ( ) Superior ( ) Superonasal ( ) Superotemporal ( ) Inferior ( ) Inferonasal ( ) Inferotemporal ( ) Cirrus OCT and RTVue findings illustrated in featured in Figures 1A and 2A, respectively. Repeatability defined as three consecutive scans taken by one operator. The resulting ICC values for assessment of intraobserver repeatability were nearly identical using one-way- or two-way random effects ANOVA. Agreement Between RTVue and Cirrus Average macular RNFL GCL-IPL thickness measured by the RTVue was lm and lm for the Cirrus OCT. Average prnfl thickness measurement was lm for RTVue and was generally thicker compared with lm measured by Cirrus. RTVue also provided a thicker prnfl measurement for the superior, inferior, nasal, and temporal quadrants (Table 5). Paired t-test revealed a highly significant difference (P < ) between each thickness measurements (average, superior, inferior, nasal, and temporal quadrant for RNFL thickness; average, superior, superonasal, superotemporal, inferior, inferonasal, and inferotemporal sector for macular RNFL-GCL-IPL thickness). Figure 3A illustrates correlation of the average prnfl between Cirrus and RTVue. Correlation of macular RNFL GCL-IPL thickness between Cirrus and RTVue is shown in Figure 3B. Values of ICC for agreement between both OCTs were generally good and ranged from in the superior segment to for the average macular RNFL GCL-IPL thickness. For the prnfl thickness, ICC ranged from in the temporal quadrant to in the inferior quadrant (Table 5). The measurements from both devices correlated well with a Pearson s correlation coefficient of (P < ) and (P < ) for average prnfl and macular RNFL GCL- IPL thickness, respectively. Quadrant and segment thickness were less correlated ranging from to for prnfl and to for macular RNFL GCL-IPL thickness. Bland-Altman plot was used to demonstrate differences in overall mean prnfl thickness measurements between RTVue and Cirrus (Fig. 4). While 94.3% of prnfl values fall within the Bland-Altman mean 6 2 SDs, there seems to be a characteristic difference between prnfl and macular RNFL GCL-IPL measurements. The values of prnfl are randomly distributed throughout the total range of thickness values while the RNFL GCL-IPL values indicate a proportional bias between Cirrus and RTVue (Fig. 5). Thickness of RNFL GCL-IPL became increasingly thicker compared to Cirrus when thickness generally, that means measured by both methods, decreased in advanced perimetric glaucoma patients with thinner inner retinal layers. In this case, a marked negative trend in the values of the differences for increasing mean values was shown indicating a TABLE 5. Macular RNFL-GCL-IPL and prnfl Thickness Cirrus, n ¼ 171 RTVue, n ¼ 171 ICC (95% CI) Macular RNFL GCL-IPL Average ( ) Superior ( ) Superonasal ( ) Superotemporal ( ) Inferior ( ) Inferonasal ( ) Inferotemporal ( ) Peripapillary RNFL Average ( ) Superior ( ) Nasal ( ) Temporal ( ) Inferior ( ) Data are mean 6 SD or as stated.

7 RNFL and Inner Retinal Thickness Using Two SD-OCTs IOVS j October 2014 j Vol. 55 j No. 10 j 6542 FIGURE 2. RTVue OCT: Raw data of macular RNFL GCL-IPL thickness calculated by the RTVue was transformed into an elliptical annulus with identical dimensions as given by the Cirrus OCT (A). Cirrus OCT: The macular RNFL thickness (inner boundary of RNFL, blue line) was added to the GCL-IPL thickness measured by the Cirrus OCT (B). proportional bias. Therefore, we considered the mean difference between the two methods in relation to the size of the measurement by regression of the difference between the measurements of both methods on the average of the measurements of the two methods (Fig. 5). DISCUSSION Our results suggest a high degree of intradevice repeatability for repeated measurements of peripapillary RNFL and macular RNFL GCL-IPL complex thickness for Cirrus and RTVue. Both OCT devices demonstrated an adequate correlation for thickness measurements, although thicknesses are different between Cirrus and RTVue. While RTVue measured a thicker prnfl with a linear relationship, measurements of the macular RNFL GCL-IPL complex became increasingly greater as the inner retinal thickness decreased in patients with thin inner retinal layers. Optical coherence tomography has become a useful tool for imaging of retinal and degenerative diseases in ophthalmology and has gained interest in neurology 7 11 as a diagnostic adjuvant to the clinical examination of MS and other neurodegenerative diseases. Recently, new developments in imaging techniques with higher image resolution have been introduced to allow segmentation of single retinal layers of the macula. RTVue was one of the first instruments capable of automatically measuring inner retinal thickness as the combined ganglion cell complex composed of macular RNFL, GCL, and IPL. 26 Cirrus OCT was the first device in which an additional automated segmentation of macular ganglion cellinner plexiform layer without RNFL was reported. 27 This approach is more challenging, because the difference in

8 RNFL and Inner Retinal Thickness Using Two SD-OCTs IOVS j October 2014 j Vol. 55 j No. 10 j 6543 FIGURE 3. Correlation of peripapillary RNFL thickness (A) and macular RNFL GCL-IPL thickness (B) between Cirrus and RTVue.

9 RNFL and Inner Retinal Thickness Using Two SD-OCTs IOVS j October 2014 j Vol. 55 j No. 10 j 6544 FIGURE 4. Bland-Altman plot to illustrate the differences in mean peripapillary RNFL thickness. Limits of agreement were provided as SD. Scatter plot with regression to illustrate the differences in mean macular RNFL GCL-IPL thickness. Limits of agreement were provided as 95% upper and lower confidence interval. FIGURE 5.

10 RNFL and Inner Retinal Thickness Using Two SD-OCTs IOVS j October 2014 j Vol. 55 j No. 10 j 6545 reflectivity between RNFL and GCL is much less than between RNFL and the vitreoretinal interface. Repeatability of scans taken by 1 or more operators on different visits should show a low intra- and interobserver variation to have reliable data. Several studies have reported on repeatability or reproducibility of different SD- or TD-OCT devices regarding peripapillary RNFL 9,13,14,16,17 and overall retinal thickness measurements. 12,15,18,22 Repeatability of inner macular thickness measurement using SD-OCT (Cirrus) has been investigated by only a few studies so far Despite the substantial differences in segmentation and acquisition algorithms between different instruments and manufacturers having their individual specifications to improve imaging techniques, correlation of thickness measurements is generally good. Results of thickness measurements are not interchangeable between SD-OCT devices and should not be used for follow-up of diseases leading to false improvement or progression of a disease. 3,9,10,12 25 This study compares the prnfl and macular ganglion cell complex (RNFL GCL-IPL) thickness using RTVue and Cirrus in healthy participants, patients with preperimetric and perimetric glaucoma, OHT patients, and patients with Parkinson disease or atypical Parkinson syndromes. We determined prnfl and macular inner retinal thickness as the potentially most sensitive parameters for ganglion cell damage in glaucoma patients and even patients with neurodegenerative diseases. Ganglion cell analyses differ considerably between Cirrus and RTVue and were therefore modified to have comparable results supporting the use of inner retinal thickness measurements. We evaluated agreement between Cirrus and RTVue and also assessed the intradevice repeatability regarding prnfl and macular RNFL GCL-IPL thickness measurements. To our knowledge, this is the first study reporting on a head-to-head comparison of inner macular thickness measurement using Cirrus and RTVue using an adjusted algorithm to have comparable measurements. Repeatability of SD-OCT (Spectralis and Cirrus) is excellent with an ICC above 0.90 and a coefficient of variation of less than 3.5%. 9,12 14 Repeatability was worse for the nasal and temporal quadrant for the RNFL measurement and the outer parts of the EDTRS sectors for the overall retinal thickness measurement. In our study, repeatability of the average macular RNFL GCL-IPL thickness was excellent with a higher ICC of (95% CI: ) for Cirrus compared with (95% CI: ) for RTVue. Repeatability was also good for average prnfl thickness measurements with an ICC of (95% CI: ) for Cirrus and (95% CI: ) for RTVue. Correlation of thickness measurements was exceptionally high for average prnfl thickness between Cirrus and RTVue in our study. Pearson s correlation coefficient was (P < ) for average prnfl thickness. Correlation for RNFL quadrants was lower ranging from to When interpreting these values, we have to bear in mind that the Pearson correlation coefficient is insensitive to scale and using it there is no account for the variance between the measurements between different raters. Intraclass coefficient correlation between Cirrus and RTVue was also good for average prnfl thickness measurement; ICC ranged from (nasal quadrant) to (inferior quadrant). The differences in correlation between quadrants is due to the arrangement of the nerve fibers in bundles with the thickest and most easily measurable RNFL in the inferior and superior quadrants. This is also in line with other studies on agreement of prnfl measurements between different OCT devices using Pearson correlation coefficient, ICC or Bland-Altman analysis. 9,13 17,23 25 RTVue generally measured a thicker prnfl thickness with linear relationship in our study. Correlation of macular RNFL GCL-IPL thickness measurements was excellent between Cirrus and RTVue in our study. Pearson correlation coefficient was (P < ) for average macular RNFL GCL-IPL thickness, although correlation for inner retinal segments was lower ranging from to Intraclass coefficient correlation between Cirrus and RTVue was in the superior segment to for the average macular RNFL GCL-IPL thickness. Despite a good correlation between OCTs, RTVue generally measured a greater macular RNFL GCL-IPL thickness. Interestingly, measurements of inner retinal thickness by the RTVue became increasingly thicker compared with Cirrus as thickness decreased in advanced glaucoma patients. This proportional bias becomes even more obvious in the Bland-Altman plots. It shows a marked negative trend in the values of the differences for increasing mean values indicating a proportional bias. Therefore, the three standard reference lines in the graphic lose their meaning in this case. It is only possible to draw conclusions in a purely descriptive manner from the plot: there is a trend and it seems that there are a roughly consistent variability across the graph. Although true inner retinal thickness is not known, we suggest that thinning of the ganglion cell and inner plexiform layer in advanced glaucoma patients may lead to a slightly false deeper placement of the outer boundary toward the inner nuclear layer in the RTVue. This phenomenon was clearly visible in a number of scans we reviewed. There are few limitations of our study. All imaging investigations were performed by a single operator and were taken three times consecutively on the same day. Therefore, our study cannot calculate interoperator or intervisit reproducibility. Moreover, in some cases both eyes of the same patient were examined ignoring the resulting intrapatient correlation in the statistical analysis. Despite these limitations, this is the first prospective study on comparison of macular ganglion cell complex measurements adjusted for differences in segmentation and thickness maps between Cirrus and RTVue. In conclusion, measurement of prnfl and macular RNFL GCL-IPL thickness has a high degree of repeatability for Cirrus and RTVue and an excellent correlation between both devices, although thickness values are substantially different for Cirrus and RTVue. RTVue generally provided a thicker prnfl measurement with a linear relationship, whereas measurements of macular RNFL GCL-IPL thickness became increasingly thicker compared with Cirrus as thickness decreased in advanced glaucoma patients with thinner inner retinal layers. Therefore, despite an excellent intradevice repeatability of scans and good correlation, devices are not interchangeable, neither for diagnosis, for follow-up, or to detect progression or illustrate improvement of a disease. Acknowledgments Supported by the Interdisciplinary Center for Clinical Research at the University of Würzburg (IZKF Würzburg, Z-2 42; JM). Disclosure: J. Matlach, IZKF Würzburg (F); M. Wagner, None; U. Malzahn, None; W. Göbel, None References 1. Huang D, Swanson EA, Lin CP, et al. Optical coherence tomography. Science. 1991;254: Adhi M, Duker JS. Optical coherence tomography current and future applications. Curr Opin Ophthalmol. 2013;24: Kiernan DF, Mieler WF, Hariprasad SM. Spectral-domain optical coherence tomography: a comparison of modern high-

11 RNFL and Inner Retinal Thickness Using Two SD-OCTs IOVS j October 2014 j Vol. 55 j No. 10 j 6546 resolution retinal imaging systems. Am J Ophthalmol. 2010; 149: Akashi A, Kanamori A, Nakamura M, Fujihara M, Yamada Y, Negi A. The ability of macular parameters and circumpapillary retinal nerve fiber layer by three SD-OCT instruments to diagnose highly myopic glaucoma. Invest Ophthalmol Vis Sci. 2013;54: Grewal DS, Tanna AP. Diagnosis of glaucoma and detection of glaucoma progression using spectral domain optical coherence tomography. Curr Opin Ophthalmol. 2013;24: Sung KR, Wollstein G, Kim NR, et al. Macular assessment using optical coherence tomography for glaucoma diagnosis. Br J Ophthalmol. 2012;96: Jindahra P, Hedges TR, Mendoza-Santiesteban CE, Plant GT. Optical coherence tomography of the retina: applications in neurology. Curr Opin Neurol. 2010;23: Watson GM, Keltner JL, Chin EK, Harvey D, Nguyen A, Park SS. Comparison of retinal nerve fiber layer and central macular thickness measurements among five different optical coherence tomography instruments in patients with multiple sclerosis and optic neuritis. J Neuroophthalmol. 2011;31: Garcia-Martin E, Pueyo V, Pinilla I, Ara JR, Martin J, Fernandez J. Fourier-domain OCT in multiple sclerosis patients: reproducibility and ability to detect retinal nerve fiber layer atrophy. Invest Ophthalmol Vis Sci. 2011;52: Garcia-Martin E, Satue M, Fuertes I, et al. Ability and reproducibility of Fourier-domain optical coherence tomography to detect retinal nerve fiber layer atrophy in Parkinson s disease. Ophthalmology. 2012;119: Marziani E, Pomati S, Ramolfo P, et al. Evaluation of retinal nerve fiber layer and ganglion cell layer thickness in Alzheimer s disease using spectral-domain optical coherence tomography. Invest Ophthalmol Vis Sci. 2013;54: Sull AC, Vuong LN, Price LL, et al. Comparison of spectral/ Fourier domain optical coherence tomography instruments for assessment of normal macular thickness. Retina. 2010;30: Tan BB, Natividad M, Chua KC, Yip LW. Comparison of retinal nerve fiber layer measurement between 2 spectral domain OCT instruments. J Glaucoma. 2012;21: Toteberg-Harms M, Sturm V, Knecht PB, Funk J, Menke MN. Repeatability of nerve fiber layer thickness measurements in patients with glaucoma and without glaucoma using spectraldomain and time-domain OCT. Graefes Arch Clin Exp Ophthalmol. 2012;250: Ho J, Sull AC, Vuong LN, et al. Assessment of artifacts and reproducibility across spectral- and time-domain optical coherence tomography devices. Ophthalmology. 2009;116: Pierro L, Gagliardi M, Iuliano L, Ambrosi A, Bandello F. Retinal nerve fiber layer thickness reproducibility using seven different OCT instruments. Invest Ophthalmol Vis Sci. 2012; 53: Seibold LK, Mandava N, Kahook MY. Comparison of retinal nerve fiber layer thickness in normal eyes using time-domain and spectral-domain optical coherence tomography. Am J Ophthalmol. 2010;150: Comyn O, Heng LZ, Ikeji F, et al. Repeatability of Spectralis OCT measurements of macular thickness and volume in diabetic macular edema. Invest Ophthalmol Vis Sci. 2012;53: Garvin MK, Lee K, Burns TL, Abramoff MD, Sonka M, Kwon YH. Reproducibility of SD-OCT-based ganglion cell-layer thickness in glaucoma using two different segmentation algorithms. Invest Ophthalmol Vis Sci. 2013;54: Hirasawa H, Araie M, Tomidokoro A, et al. Reproducibility of thickness measurements of macular inner retinal layers using SD-OCT with or without correction of ocular rotation. Invest Ophthalmol Vis Sci. 2013;54: Francoz M, Fenolland JR, Giraud JM, et al. Reproducibility of macular ganglion cell-inner plexiform layer thickness measurement with cirrus HD-OCT in normal, hypertensive and glaucomatous eyes. Br J Ophthalmol. 2014;98: Wolf-Schnurrbusch UE, Ceklic L, Brinkmann CK, et al. Macular thickness measurements in healthy eyes using six different optical coherence tomography instruments. Invest Ophthalmol Vis Sci. 2009;50: Kanamori A, Nakamura M, Tomioka M, Kawaka Y, Yamada Y, Negi A. Agreement among three types of spectral-domain optical coherent tomography instruments in measuring parapapillary retinal nerve fibre layer thickness. Br J Ophthalmol. 2012;96: Lee ES, Kang SY, Choi EH, et al. Comparisons of nerve fiber layer thickness measurements between Stratus, Cirrus, and RTVue OCTs in healthy and glaucomatous eyes. Optom Vis Sci. 2011;88: Leite MT, Rao HL, Weinreb RN, et al. Agreement among spectral-domain optical coherence tomography instruments for assessing retinal nerve fiber layer thickness. Am J Ophthalmol. 2011;151:85 92.e Tan O, Chopra V, Lu AT, et al. Detection of macular ganglion cell loss in glaucoma by Fourier-domain optical coherence tomography. Ophthalmology. 2009;116: e Mwanza JC, Oakley JD, Budenz DL, Chang RT, Knight OJ, Feuer WJ. Macular ganglion cell-inner plexiform layer: automated detection and thickness reproducibility with spectral domain-optical coherence tomography in glaucoma. Invest Ophthalmol Vis Sci. 2011;52: