[Skip to Content]
Access to paid content on this site is currently suspended due to excessive activity being detected from your IP address 54.159.202.12. Please contact the publisher to request reinstatement.
[Skip to Content Landing]
Download PDF
Figure 1.
Difference of Stratus Optical Coherence Tomography (OCT) and Spectralis OCT Mean Retinal Nerve Fiber Layer Thickness (RFNLT)
Difference of Stratus Optical Coherence Tomography (OCT) and Spectralis OCT Mean Retinal Nerve Fiber Layer Thickness (RFNLT)

Mean difference RNFLT measurements between Stratus OCT and Spectralis OCT, with 95% CI RFNLT for each diagnostic group (A) and the equation and test groups (B). Error bars indicate 95% CIs.

aSignificant difference of P < .05.

Figure 2.
Bland-Altman Plots Assess the Agreement Between Predicted and Measured Mean Retinal Nerve Fiber Layer Thickness (RNFLT)
Bland-Altman Plots Assess the Agreement Between Predicted and Measured Mean Retinal Nerve Fiber Layer Thickness (RNFLT)

The difference between predicted and measured Stratus OCT (A-D) and Spectralis OCT (E-H) RNFLT in control participants, patients with ocular hypertension, patients with preperimetric glaucoma, and patients with perimetric glaucoma.

Figure 3.
Influence of Glaucoma Stage on the Intraclass Correlation Coefficient (ICC)
Influence of Glaucoma Stage on the Intraclass Correlation Coefficient (ICC)

Spectralis optical coherence tomography (OCT) and Stratus OCT unconverted peripapillary retinal nerve fiber layer thickness measurements were evaluated in 32 peripapillary sectors, demonstrating significantly lower ICCs in 2 nasoinferior sectors for patients with preperimetric glaucoma and patients with perimetric glaucoma compared with control participants. Error bars indicate 95% CIs.

aSignificantly lower intraclass correlation coefficient (P < .05).

Table 1.  
Demographic and Clinical Characteristics of Included Participantsa
Demographic and Clinical Characteristics of Included Participantsa
Table 2.  
Demographic and Clinical Characteristics of Included Participants in the Equation and Test Groupsa
Demographic and Clinical Characteristics of Included Participants in the Equation and Test Groupsa
Table 3.  
Equations Derived From the Equation Group and Applied to the Test Group to Calculate the Predicted Mean RNFLT
Equations Derived From the Equation Group and Applied to the Test Group to Calculate the Predicted Mean RNFLT
1.
Huang  D, Swanson  EA, Lin  CP,  et al.  Optical coherence tomography. Science. 1991;254(5035):1178-1181.
PubMedArticle
2.
Wollstein  G, Schuman  JS, Price  LL,  et al.  Optical coherence tomography (OCT) macular and peripapillary retinal nerve fiber layer measurements and automated visual fields. Am J Ophthalmol. 2004;138(2):218-225.
PubMedArticle
3.
Sihota  R, Sony  P, Gupta  V, Dada  T, Singh  R.  Diagnostic capability of optical coherence tomography in evaluating the degree of glaucomatous retinal nerve fiber damage. Invest Ophthalmol Vis Sci. 2006;47(5):2006-2010.
PubMedArticle
4.
Wollstein  G, Schuman  JS, Price  LL,  et al.  Optical coherence tomography longitudinal evaluation of retinal nerve fiber layer thickness in glaucoma. Arch Ophthalmol. 2005;123(4):464-470.
PubMedArticle
5.
Deleón-Ortega  JE, Arthur  SN, McGwin  G  Jr, Xie  A, Monheit  BE, Girkin  CA.  Discrimination between glaucomatous and nonglaucomatous eyes using quantitative imaging devices and subjective optic nerve head assessment. Invest Ophthalmol Vis Sci. 2006;47(8):3374-3380.
PubMedArticle
6.
Wessel  JM, Horn  FK, Tornow  RP,  et al.  Longitudinal analysis of progression in glaucoma using spectral-domain optical coherence tomography. Invest Ophthalmol Vis Sci. 2013;54(5):3613-3620.
PubMedArticle
7.
Miki  A, Medeiros  FA, Weinreb  RN,  et al.  Rates of retinal nerve fiber layer thinning in glaucoma suspect eyes. Ophthalmology. 2014;121(7):1350-1358.
PubMedArticle
8.
Schuman  JS.  Spectral domain optical coherence tomography for glaucoma (an AOS thesis). Trans Am Ophthalmol Soc. 2008;106:426-458.
PubMed
9.
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(9):5912-5920.
PubMedArticle
10.
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(6):807-814.
PubMedArticle
11.
Larrosa  JM, Polo  V, Ferreras  A, García-Martín  E, Calvo  P, Pablo  LE.  Neural network analysis of different segmentation strategies of nerve fiber layer assessment for glaucoma diagnosis. J Glaucoma. 2014;(Jul):22.
PubMed
12.
Matsumoto  M, Nishimura  T.  Mersenne twister: a 623-dimensionally equidistributed uniform pseudo-random number generator. ACM Trans Model Comput Simul. 1998;8(1):3-30.Article
13.
Bland  JM, Altman  DG.  Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1(8476):307-310.
PubMedArticle
14.
Rebolleda  G, González-López  JJ, Muñoz-Negrete  FJ, Oblanca  N, Costa-Frossard  L, Álvarez-Cermeño  JC.  Color-code agreement among stratus, cirrus, and spectralis optical coherence tomography in relapsing-remitting multiple sclerosis with and without prior optic neuritis. Am J Ophthalmol. 2013;155(5):890-897.
PubMedArticle
15.
Armitage  P, Berry  G, Matthews  JNS. Statistical Methods in Medical Research. Hoboken, NJ: Wiley; 2008.
16.
Vizzeri  G, Weinreb  RN, Gonzalez-Garcia  AO,  et al.  Agreement between spectral-domain and time-domain OCT for measuring RNFL thickness. Br J Ophthalmol. 2009;93(6):775-781.
PubMedArticle
17.
González-García  AO, Vizzeri  G, Bowd  C, Medeiros  FA, Zangwill  LM, Weinreb  RN.  Reproducibility of RTVue retinal nerve fiber layer thickness and optic disc measurements and agreement with Stratus optical coherence tomography measurements. Am J Ophthalmol. 2009;147(6):1067-1074, 1074.e1.
PubMedArticle
18.
Leung  CK, Chiu  V, Weinreb  RN,  et al.  Evaluation of retinal nerve fiber layer progression in glaucoma: a comparison between spectral-domain and time-domain optical coherence tomography. Ophthalmology. 2011;118(8):1558-1562.
PubMedArticle
19.
Leung  CK, Cheung  CY, Weinreb  RN,  et al.  Retinal nerve fiber layer imaging with spectral-domain optical coherence tomography: a variability and diagnostic performance study. Ophthalmology. 2009;116(7):1257-1263, 1263.e1-1263.e2.
PubMedArticle
20.
Töteberg-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 spectral-domain and time-domain OCT. Graefes Arch Clin Exp Ophthalmol. 2012;250(2):279-287.
PubMedArticle
21.
Brito  PN, Vieira  MP, Falcão  MS, Faria  OS, Falcão-Reis  F.  Optical coherence tomography study of peripapillary retinal nerve fiber layer and choroidal thickness in eyes with tilted optic disc. J Glaucoma. 2015;24(1):45-50.
PubMedArticle
22.
Pinto  LM, Costa  EF, Melo  LA  Jr,  et al.  Structure-function correlations in glaucoma using matrix and standard automated perimetry versus time-domain and spectral-domain OCT devices. Invest Ophthalmol Vis Sci. 2014;55(5):3074-3080.
PubMedArticle
23.
Nukada  M, Hangai  M, Mori  S,  et al.  Detection of localized retinal nerve fiber layer defects in glaucoma using enhanced spectral-domain optical coherence tomography. Ophthalmology. 2011;118(6):1038-1048.
PubMedArticle
24.
Nukada  M, Hangai  M, Mori  S,  et al.  Imaging of localized retinal nerve fiber layer defects in preperimetric glaucoma using spectral-domain optical coherence tomography. J Glaucoma. 2014;23(3):150-159.
PubMedArticle
25.
Shin  HJ, Cho  BJ.  Comparison of retinal nerve fiber layer thickness between Stratus and Spectralis OCT. Korean J Ophthalmol. 2011;25(3):166-173.
PubMedArticle
26.
Knighton  RW, Qian  C.  An optical model of the human retinal nerve fiber layer: implications of directional reflectance for variability of clinical measurements. J Glaucoma. 2000;9(1):56-62.
PubMedArticle
27.
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(4):266-273.
PubMedArticle
28.
Budenz  DL, Fredette  MJ, Feuer  WJ, Anderson  DR.  Reproducibility of peripapillary retinal nerve fiber thickness measurements with stratus OCT in glaucomatous eyes. Ophthalmology. 2008;115(4):661-666.e4, e664.
PubMedArticle
29.
Leung  CK, Mohamed  S, Leung  KS,  et al.  Retinal nerve fiber layer measurements in myopia: an optical coherence tomography study. Invest Ophthalmol Vis Sci. 2006;47(12):5171-5176.
PubMedArticle
30.
Hwang  YH, Lee  JY, Kim  YY.  The effect of head tilt on the measurements of retinal nerve fibre layer and macular thickness by spectral-domain optical coherence tomography. Br J Ophthalmol. 2011;95(11):1547-1551.
PubMedArticle
31.
Kim  JS, Ishikawa  H, Gabriele  ML,  et al.  Retinal nerve fiber layer thickness measurement comparability between time domain optical coherence tomography (OCT) and spectral domain OCT. Invest Ophthalmol Vis Sci. 2010;51(2):896-902.
PubMedArticle
Original Investigation
October 2015

Predicted and Measured Retinal Nerve Fiber Layer Thickness From Time-Domain Optical Coherence Tomography Compared With Spectral-Domain Optical Coherence Tomography

Author Affiliations
  • 1Department of Ophthalmology, University of Erlangen–Nuremberg, Erlangen, Germany
JAMA Ophthalmol. 2015;133(10):1135-1143. doi:10.1001/jamaophthalmol.2015.2427
Abstract

Importance  New methods are needed to compare peripapillary retinal nerve fiber layer thickness (pRNFLT) measurements taken from time-domain optical coherence tomography (TD-OCT) and spectral-domain OCT (SD-OCT).

Objective  To compare the agreement of measured and predicted pRNFLT using different equations based on pRNFLT measurements obtained by TD-OCT and SD-OCT.

Design, Setting, and Participants  Cross-sectional single-center study that took place at the Department of Ophthalmology, University of Erlangen–Nuremberg from November 16, 2005, to June 3, 2015, and included 138 eyes of control participants, 126 eyes of patients with ocular hypertension, 128 eyes of patients with preperimetric glaucoma, and 160 eyes of patients with perimetric glaucoma. All participants had standard clinical examinations to obtain TD-OCT (via Stratus OCT) and SD-OCT (via Spectralis OCT) measurements of pRNFLT. Two groups were matched for diagnostic subgroup, eye side, sex, and age. The TD-OCT measurements of the first group were used to predict the mean SD-OCT and 6-sector vertical-split pRNFLT measurements of the second group and vice versa. The agreement between the predicted pRNFLT calculations of conversion equations and measured pRNFLT of the second group was evaluated by intraclass correlation coefficients and Bland-Altman plots.

Main Outcomes and Measures  Mean and sectoral pRNFLT measurements obtained by TD-OCT and SD-OCT as well as the agreement between measured and predicted pRNFLT.

Results  The agreement for all investigated equations to predict mean pRNFLT measurements with intraclass correlation coeffecients ranged from 0.937 to 0.939. Bland-Altman plots demonstrated systemic biases between −0.7 μm and +1.1 μm for measured and predicted mean pRNFLT measurements. The ratio method demonstrated an intraclass correlation coefficient of 0.969 for the temporal-inferior sector. The best color-code agreement between both OCT devices was achieved by the no conversion method, with κ = 0.731 (95% CI, 0.656-0.806) for the mean pRNFLT.

Conclusions and Relevance  These data suggest that the prediction of mean pRNFLT values by equations derived from TD-OCT and SD-OCT can be conducted with high levels of agreement. In individual cases and singular sectors, high prediction errors may occur. When longitudinal imaging data from both TD-OCT and SD-OCT are available, conversion equations may provide longitudinal comparability.

Introduction

Optical coherence tomography (OCT) has become a rapidly evolving imaging technique since its introduction in 1991.1 It has led to significant enhancements of our knowledge of early detection and monitoring of optic nerve disorders, such as glaucoma.

The assessment of retinal nerve fiber layer thickness (RNFLT) obtained by OCT is a key diagnostic feature of structural glaucomatous damage that has been shown by numerous cross-sectional25 and longitudinal studies.4,6,7 For many years, time-domain optical coherence tomography (TD-OCT) was widely used in the diagnosis and follow-up of patients with glaucoma. The evolution from TD-OCT to spectral-domain OCT (SD-OCT) led to a more than 100-fold faster acquisition process, increased axial image resolution,8 and enabled the development of eye tracking systems and noise-reduction techniques, leading to a higher repeatability of measurements.

A method to compare peripapillary retinal nerve fiber layer thickness (pRNFLT) measurements obtained from TD-OCT and SD-OCT in patients who were imaged first by one type of OCT and then the other would facilitate analysis of the progression of glaucomatous optic neuropathy. Furthermore, it would allow comparing pRNFLT obtained by TD-OCT devices reported in earlier clinical trials with those obtained by SD-OCT devices in more recent investigations. Previously, Pierro et al9 and Seibold et al10 performed a comprehensive comparison between TD-OCT and SD-OCT pRNFLT measurements, including 3 or more different SD-OCT devices; however, those studies included only healthy eyes.

The purpose of the present study was to investigate whether pRNFLT measurements of SD-OCT using Spectralis OCT could be reliably predicted by means of equations based on TD-OCT measurements using Stratus OCT and vice versa and to compare the agreement and accuracy of measured and predicted pRNFLT.

Box Section Ref ID

At a Glance

  • The main goal study of this study is to bridge retinal nerve fiber layer measurements of time-domain optical coherence tomography and spectral-domain optical coherence tomography.

  • The comparability of time-domain optical coherence tomography and spectral-domain optical coherence tomography mean retinal nerve fiber layer measurements is feasible for group comparisons.

  • Conversion equations may not be necessary to bridge retinal nerve fiber layer measurements of time-domain optical coherence tomography and spectral-domain optical coherence tomography.

  • The achievement of longer follow-up data will allow for an analysis of long-term effects and the progression of glaucomatous optic neuropathy.

Methods
Study Design

This cross-sectional controlled single-center study was performed in accordance with the tenets of the Declaration of Helsinki for research involving human participants and was approved by the ethics review board of the medical faculty of the University of Erlangen-Nuremberg. Participants of the study were selected from the Erlangen Glaucoma Registry (1460 participants). We obtained written informed consent from all participants.

If both eyes of a participant were eligible for this study, 1 eye was selected at random. Patients and control participants were examined comprehensively by slitlamp biomicroscopy, gonioscopy, and dilated funduscopy. Goldmann applanation tonometry was performed 5 times in 24 hours to assess intraocular pressure in all patients.

Papillomorphologic evaluation was based on telecentric fundus camera 15° optic disc slides (Carl Zeiss Meditec). Appearance of the optic nerve head was classified as either healthy or glaucomatous based on the absence or presence of glaucomatous optic disc damage characterized by focal rim notching, diffuse or localized wedge-shaped defects of the RNFL, and an abnormally small neuroretinal rim area, with respect to the size of the optic nerve head and optic disc hemorrhages. Two masked glaucoma specialists (C.Y.M. and R.L.) independently graded optic disc slides. In cases of discrepancy, a third clinical glaucoma specialist who was unaware of the diagnosis made the judgment.

Participants with ocular hypertension had intraocular pressure levels more than 21 mm Hg and no signs of structural or functional glaucomatous damage. Patients with preperimetric glaucoma had signs of optic disc damage on optic disc slides without any corresponding visual field loss. Patients with perimetric glaucoma had a glaucomatous-appearing optic disc and pathologically cumulative perimetric defect curves.

All participants included in this study had a best-corrected visual acuity of 20/30 Snellen equivalents or better and clear optic media. Participants were excluded if they had a myopic refractive error exceeding −8 diopters of sphere, a hyperopic refractive error exceeding +6 diopters or 4 diopters of astigmatism, the presence of diabetes mellitus, or any eye disease other than primary open-angle glaucoma.

Spectral-Domain Optical Coherence Tomography

The pRNFLT measurements were obtained by the Spectralis HRA+OCT SD-OCT system (Heidelberg Engineering). Infrared reflection images (λ = 820 nm) and OCT B-scans (λ = 870 nm, 40 000 A-scans per second with 7-µm axial optical resolution) of the dual-beam laser scanning systems were acquired concurrently. The mean of 16 successive circular B-scans (768 A-scans with 12° diameter) centered at the optic disc was automatically calculated to reduce speckle noise. During image acquisition, an incorporated eye tracking system compensated for eye movements. An experienced operator controlled and corrected all segmentations of RNFLT as previously described.6 Spectralis OCT, version 5.3, was used. The mean pRNFLT and 6 vertical-split sectors of pRNFLT as described by Larrosa et al11 were evaluated.

Time-Domain Optical Coherence Tomography

In this study, TD-OCT imaging with Stratus OCT, version 4.0.1 (Carl Zeiss Meditec Inc), was performed. The regular pRNFLT scan protocol with a 3.46-mm circular scan was used to measure pRNFLT (512 A-scans). The mean of 3 consecutive circular scans was evaluated.

Imaging Protocol and Quality Control

An operator who was masked from the scan data obtained all measurements of pRNFLT with both imaging devices in a single session. The TD-OCT and SD-OCT scans were performed in a randomly assigned order on the same day immediately after each other. For TD-OCT only, scans with signal strengths greater than or equal to 5 (maximum, 10) were accepted for further analysis. To decrease the effect of motion artifacts known to occur in TD-OCT, we calculated the mean of 3 single peripapillary scans. For both SD-OCT and TD-OCT, scans with displacement of the scan circle as well as scans with mirror, blink, and out-of-register artifacts were excluded from further analysis. From 1690 TD-OCT scans, 20 (1.2%) were excluded owing to low signal strength and 14 (0.8%) were excluded owing to artifacts or displacement of the scan circle. From 555 SD-OCT scans, 3 (0.5%) were rejected owing to artifacts. A total of 1656 scans for TD-OCT and 552 scans for SD-OCT were included in the final analysis. To compare Stratus OCT measurements with Spectralis OCT measurements, we calculated the mean thickness data of 32 peripapillary sectors (11.25° each). To compare the predicted and actual values of sector RNFLT, we formed 6 new vertical-split sectors for Stratus OCT according to the Spectralis OCT sectors by calculating the mean.

Perimetry

All patients and control participants were measured with standard white-on-white automated perimetry by means of an Octopus 500 computerized automatic static perimeter (Haag-Streit). The percentage of false-negative and false-positive answers had to be lower than 15%. A pathologic visual field test was characterized by clusters of at least 3 adjacent points with P < .05 or 2 adjacent points with P < .01 in the pattern deviation map assessed with Peridata software, version 2.2.3. The first 2 visual field tests from each patient were excluded from further analysis to avoid learning effects.

Main Outcomes and Measures

The cohort was divided into 2 equally composed groups of patients matched for diagnostic subgroup, eye side, sex, and age by a block pseudorandomization process that used the Mersenne-Twister algorithm.12 The OCT data of the first group were used to determine equations (equation group), which were then tested on the second group (test group).

Stratus OCT measurements of the first group were used to predict the pRNFLT of the Spectralis OCT measurements and vice versa. Equations for the mean and 6-sector vertical-split pRNFLT measurement were constructed by the following methods: no conversion, mean difference, multiplication of a mean ratio, and linear regression.

For the no conversion method, we directly used the pRNFLT of the Spectralis OCT measurement to predict the Stratus OCT measurement and vice versa; thus, mean pRNFLT was used interchangeably.

For the mean difference method, we calculated the difference between Spectralis OCT and Stratus OCT for mean and sector data. The difference was subtracted from the Stratus OCT pRNFLT of each participant to predict Spectralis OCT pRNFLT and vice versa.

The mean ratio method involved computing the mean ratio of pRNFLT measurements from Spectralis OCT and Stratus OCT for mean and sector data. The computed value was multiplied by the Stratus OCT pRNFLT of each participant to predict the Spectralis OCT pRNFLT and vice versa.

Linear regression analysis was conducted to fit predictive linear models (y = ax + b) to the observed mean and sector pRNFLT measurements, where y represented the predicted value of 1 OCT, a represented the slope of the regression line, x represented the measured value of the other OCT, and b represented the y-intercept. These fitted models were used to make a prediction of the individual pRNFLT for mean and sector data.

To evaluate the intraclass repeatability of predicted and actual pRNFLT, we calculated the intraclass correlation coefficient (ICC) and 95% CI for mean and sector OCT data. For the ICC calculation, the 2-way single-ratings mixed-effects model for the measure of absolute agreement was used. Agreement was assessed by Bland-Altman plots13 between the predicted and measured pRNFLT.

The pRNFLT color-code agreement (green, yellow, red, and white) as previously described by Rebolleda et al14 was evaluated by the Cohen κ for mean data after applying the 4 methods mentioned earlier. Mean pRNFLT measurements of all control participants included in this study were used to obtain the age-adjusted first percentile, fifth percentile, and 95th percentile values.

To assess potential demographic and clinical differences between diagnostic subgroups and between test and equation groups with respect to quantitative variables, we used 1-way analysis of variance and paired t tests. For differences in qualitative variables between subgroups, we performed the Pearson χ2 test. To counteract for the inflation of type I error owing to multiple comparisons, we used the Bonferroni correction.15 Statistical analysis was conducted using SPSS software, version 22.0 (IBM Corp). The significance level was P < .05.

Results

A total of 138 eyes of control participants, 126 eyes of patients with ocular hypertension, 128 eyes of patients with preperimetric glaucoma, and 160 eyes of patients with perimetric glaucoma were imaged by Stratus OCT and Spectralis OCT. Patients’ characteristics and statistical results are summarized in Tables 1 and 2. The mean defect and the square root of the loss of variance were significantly worse in the perimetric glaucoma group compared with all other diagnostic groups (P < .001).

The control participants were younger than all other groups and the perimetric glaucoma group was significantly older than the ocular hypertension and control groups. The equation and test groups did not differ in age, sex, eye side, diagnostic group, mean defect, and square root of loss of variance. No significant differences between the pRNFLT measurements of both imaging devices were observed for the mean pRNFLT of each diagnostic group and the test and equation groups (Figure 1). There were significant differences between sector pRNFLT measurements in all groups except for the preperimetric glaucoma group (Figure 1). Underlying pRNFLT data in all groups and sectors are summarized in eTable 1 in the Supplement. Equations that were derived from OCT measurements of the equation group are displayed in Table 3.

The percentage of data within 95% of limits of agreement was between 94.6% and 95.3% for the predicted Spectralis OCT and measured Spectralis OCT mean pRNFLT (Figure 2) and 94.9% and 95.3% for the predicted Stratus OCT and measured Stratus OCT mean pRNFLT (Figure 2). Systemic biases were between −0.7 and +1.1 µm and no proportional biases were observed (Figure 2).

For mean pRNFLT, ICCs ranged between 0.937 and 0.939 for all equation methods (eTable 2 in the Supplement). The regression equation method achieved the highest percentage of eyes within a difference smaller than or equal to 5% between the actual and predicted RNFLT, with 51.1% to predict Spectralis OCT mean pRNFLT and 53.6% to predict Stratus OCT mean pRNFLT (eTable 3 in the Supplement). The ratio method showed the highest ICC values for all regional sectors; however, those were not statistically significantly different from the values obtained by unconverted prediction (eTable 4 in the Supplement). The ICCs ranged between 0.770 for the nasal sector and 0.969 for the temporal-inferior sector to predict the Spectralis OCT pRNFLT and between 0.800 and 0.968 to predict the Stratus OCT pRNFLT. For patients with preperimetric glaucoma and perimetric glaucoma, the ICCs for unconverted prediction were lower compared with control participants in 2 nasoinferior sectors (Figure 3).

Color-code agreement between the predicted and actual mean Stratus OCT and Spectralis OCT ranged between κ = 0.678 (95% CI, 0.599-0.757) and κ = 0.709 (95% CI, 0.632-0.777), respectively, for the mean difference method and κ = 0.731 (95% CI, 0.656-0.806) for the no conversion method.

Discussion

The transition from TD-OCT to SD-OCT imaging to document pRNFLT and monitor the structural progression of glaucoma leads to questions on the agreement and predictability of pRNFLT between the 2 different OCT technologies. In this study, we demonstrated that pRNFLT obtained by Spectralis OCT and Stratus OCT devices show high agreement and good predictability for mean pRNFLT measurements.

To date, different studies have addressed comparisons of pRNFLT measurements of SD-OCT and TD-OCT devices.9,1619 Prior studies compared Spectralis OCT and Stratus OCT pRNFLT measurements by answering different scientific questions.2025 Töteberg-Harms et al20 found the intra- and interobserver repeatability of Spectralis OCT superior to Stratus OCT, especially for glaucomatous eyes. Pinto et al22 described strong structure-function correlations between the visual fields and both Spectralis OCT and Stratus OCT pRNFLT measurements in the overall and arcuate sectors. Nukada et al23,24 showed Spectralis OCT to be superior to Stratus OCT to detect wedge-shaped localized pRNFLT defects as seen on red-free fundus photographs for patients with perimetric and preperimetric glaucoma. Good agreement and high correlation between Spectralis OCT and Stratus OCT was demonstrated previously.9,10,25

In our study, the raw OCT data of both devices were comparable in most of the investigated sectors and subgroups. However, in accordance with previous investigators,10,25 we found absolute differences between Stratus OCT and Spectralis OCT pRNFLT measurements. The Spectralis OCT showed significantly higher values in 3 sectors of 3 of the diagnostic subgroups compared with Stratus OCT and significantly lower values in 3 sectors of 2 diagnostic subgroups. These deviations might be owing to differences in segmentation algorithms of both devices or in image-processing procedures, such as the smoothing algorithm of Stratus OCT. Moreover, pRNFLT segmentation errors could be corrected manually for Spectralis OCT but not for Stratus OCT. Therefore, the segmentation errors of Stratus OCT that could not be corrected might have caused differences in pRNFLT measurements. In contrast to Shin et al,25 we found no differences in the mean pRNFLT. This might be related to the higher number of participants included in our study.

We observed a significant decrease of the ICC in the preperimetric and perimetric glaucoma groups compared with the control group in 2 nasal sectors. This might be related to the deteriorating signal to noise ratio for both imaging devices in these sectors and/or dimmer illumination of the nasal RNFL because of the directional reflectance properties of the RNFL, previously reported by Knighton et al.26

We could not observe significant differences between the best method and unconverted prediction for group comparisons. Bland-Altman plots showed a comparable percentage of data points within the 95% limits of agreements for all investigated methods and the CIs of the ICCs showed an overlap of the 95% CI. This may be owing to the small mean difference of pRNFLT measured by Spectralis OCT and Stratus OCT.

A test-retest variability of 4.95 µm for mean RNFLT in healthy eyes for Spectralis OCT was reported by Tan et al27 and 5.2 µm in glaucomatous eyes for Stratus OCT was reported by Budenz.28

For all investigated methods, including no conversion, roughly half the eyes in our study showed a difference smaller than 5% between the measured and predicted mean RNFLT for both devices, which, presumably, is in the range of test-retest variability of Spectralis OCT and Stratus OCT. In our study, we included myopic patients with a myopic refractive error of as many as −8 diopters. Because highly myopic eyes have been shown to demonstrate significantly lower pRNFLT than low-moderate myopic eyes,29 one may speculate that including these patients in our study might have deteriorated the agreement of both OCT devices owing to a smaller signal to noise ratio or other yet unknown factors in these patients.

To predict sector pRNFLT, the ratio method showed the highest accuracy. However, CIs of sector conversion factors showed a considerable overlap, suggesting that a conversion equation may not be necessary for the prediction of sector pRNFLT. The prediction of mean pRNFLT was more robust in comparison with sector pRNFLT. This may be owing to the lower SD of mean pRNFLT compared with sector pRNFLT or owing to differences in the torsional eye position during scanning by Stratus OCT and Spectralis OCT. Particularly, head tilt might produce a significant displacement of the temporal superior nasal inferior temporal plot, with little effect on the global pRNFLT.30 Because TD-OCT and SD-OCT imaging could not be performed simultaneously, a possible limitation of the direct comparison of our study was the operator-dependent positioning of the scan circle and possible scattering along the scan circle owing to eye motion.31

By linking longitudinal pRNFLT measurements obtained by Stratus OCT and Spectralis OCT to achieve longer follow-up data, new insights into the long-term effects and progression of glaucomatous optic neuropathy can be gained, which is of high value owing to the slowly progressive nature of the disease.

We propose that the methods reported in this study could easily be applied to different study populations when longitudinal imaging data from both Stratus OCT and Spectralis OCT are available to determine the best method for conversion and provide longitudinal comparability. As proposed by Kim et al,31 to reduce variability in 3-dimensional volume data of SD-OCT when compared with TD-OCT, a scan location-matching algorithm could be applied.

Conclusions

Given the lack of confirmatory studies and based on our data, we believe that group comparisons in clinical research could be performed and that the longitudinal comparability of Stratus OCT and Spectralis OCT mean pRNFLT measurements is feasible for OCT data with substantial color-code agreement and excellent correlation for mean temporal-superior and temporal-inferior sectors. Although we identified equations that could ameliorate the agreement of Stratus OCT and Spectralis OCT pRNFLT measurements, 95% CIs of all equations except the regression equation method (eTables 3 and 4 in the Supplement) included 1 for the slope and 0 for the y-intercept, indicating that conversion equations may not be necessary to provide comparability of both imaging devices.

We do not recommend using pRNFLT measurement data of both OCTs interchangeably for treatment decisions of individual patients in clinical practice because this may lead to the misjudgment of pRNFLT owing to possible high prediction errors.

Back to top
Article Information

Corresponding Author: Wolfgang A. Schrems, MD, Department of Ophthalmology, University Eye Hospital, Schwabachanlage 6, 91054 Erlangen, Germany (wolfgang.schrems@uk-erlangen.de).

Submitted for Publication: February 24, 2015; final revision received June 3, 2015; accepted June 11, 2015.

Published Online: July 30, 2015. doi:10.1001/jamaophthalmol.2015.2427.

Author Contributions: Dr Schrems and Schrems-Hoesl had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Drs Schrems and Schrems-Hoesl contributed equally to this work.

Study concept and design: All authors.

Acquisition, analysis, or interpretation of data: Schrems, Schrems-Hoesl, Bendschneider, Mardin, Laemmer, Horn.

Drafting of the manuscript: Schrems, Schrems-Hoesl.

Critical revision of the manuscript for important intellectual content: Bendschneider, Mardin, Laemmer, Kruse, Horn.

Statistical analysis: Schrems, Schrems-Hoesl, Horn.

Obtained funding: Mardin.

Administrative, technical, or material support: Mardin, Horn.

Study supervision: Bendschneider, Mardin, Laemmer, Kruse.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.

Funding/Support: The study was supported in parts by grant SFB539 from the German Research Foundation from 1997 to 2009.

Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

References
1.
Huang  D, Swanson  EA, Lin  CP,  et al.  Optical coherence tomography. Science. 1991;254(5035):1178-1181.
PubMedArticle
2.
Wollstein  G, Schuman  JS, Price  LL,  et al.  Optical coherence tomography (OCT) macular and peripapillary retinal nerve fiber layer measurements and automated visual fields. Am J Ophthalmol. 2004;138(2):218-225.
PubMedArticle
3.
Sihota  R, Sony  P, Gupta  V, Dada  T, Singh  R.  Diagnostic capability of optical coherence tomography in evaluating the degree of glaucomatous retinal nerve fiber damage. Invest Ophthalmol Vis Sci. 2006;47(5):2006-2010.
PubMedArticle
4.
Wollstein  G, Schuman  JS, Price  LL,  et al.  Optical coherence tomography longitudinal evaluation of retinal nerve fiber layer thickness in glaucoma. Arch Ophthalmol. 2005;123(4):464-470.
PubMedArticle
5.
Deleón-Ortega  JE, Arthur  SN, McGwin  G  Jr, Xie  A, Monheit  BE, Girkin  CA.  Discrimination between glaucomatous and nonglaucomatous eyes using quantitative imaging devices and subjective optic nerve head assessment. Invest Ophthalmol Vis Sci. 2006;47(8):3374-3380.
PubMedArticle
6.
Wessel  JM, Horn  FK, Tornow  RP,  et al.  Longitudinal analysis of progression in glaucoma using spectral-domain optical coherence tomography. Invest Ophthalmol Vis Sci. 2013;54(5):3613-3620.
PubMedArticle
7.
Miki  A, Medeiros  FA, Weinreb  RN,  et al.  Rates of retinal nerve fiber layer thinning in glaucoma suspect eyes. Ophthalmology. 2014;121(7):1350-1358.
PubMedArticle
8.
Schuman  JS.  Spectral domain optical coherence tomography for glaucoma (an AOS thesis). Trans Am Ophthalmol Soc. 2008;106:426-458.
PubMed
9.
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(9):5912-5920.
PubMedArticle
10.
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(6):807-814.
PubMedArticle
11.
Larrosa  JM, Polo  V, Ferreras  A, García-Martín  E, Calvo  P, Pablo  LE.  Neural network analysis of different segmentation strategies of nerve fiber layer assessment for glaucoma diagnosis. J Glaucoma. 2014;(Jul):22.
PubMed
12.
Matsumoto  M, Nishimura  T.  Mersenne twister: a 623-dimensionally equidistributed uniform pseudo-random number generator. ACM Trans Model Comput Simul. 1998;8(1):3-30.Article
13.
Bland  JM, Altman  DG.  Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1(8476):307-310.
PubMedArticle
14.
Rebolleda  G, González-López  JJ, Muñoz-Negrete  FJ, Oblanca  N, Costa-Frossard  L, Álvarez-Cermeño  JC.  Color-code agreement among stratus, cirrus, and spectralis optical coherence tomography in relapsing-remitting multiple sclerosis with and without prior optic neuritis. Am J Ophthalmol. 2013;155(5):890-897.
PubMedArticle
15.
Armitage  P, Berry  G, Matthews  JNS. Statistical Methods in Medical Research. Hoboken, NJ: Wiley; 2008.
16.
Vizzeri  G, Weinreb  RN, Gonzalez-Garcia  AO,  et al.  Agreement between spectral-domain and time-domain OCT for measuring RNFL thickness. Br J Ophthalmol. 2009;93(6):775-781.
PubMedArticle
17.
González-García  AO, Vizzeri  G, Bowd  C, Medeiros  FA, Zangwill  LM, Weinreb  RN.  Reproducibility of RTVue retinal nerve fiber layer thickness and optic disc measurements and agreement with Stratus optical coherence tomography measurements. Am J Ophthalmol. 2009;147(6):1067-1074, 1074.e1.
PubMedArticle
18.
Leung  CK, Chiu  V, Weinreb  RN,  et al.  Evaluation of retinal nerve fiber layer progression in glaucoma: a comparison between spectral-domain and time-domain optical coherence tomography. Ophthalmology. 2011;118(8):1558-1562.
PubMedArticle
19.
Leung  CK, Cheung  CY, Weinreb  RN,  et al.  Retinal nerve fiber layer imaging with spectral-domain optical coherence tomography: a variability and diagnostic performance study. Ophthalmology. 2009;116(7):1257-1263, 1263.e1-1263.e2.
PubMedArticle
20.
Töteberg-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 spectral-domain and time-domain OCT. Graefes Arch Clin Exp Ophthalmol. 2012;250(2):279-287.
PubMedArticle
21.
Brito  PN, Vieira  MP, Falcão  MS, Faria  OS, Falcão-Reis  F.  Optical coherence tomography study of peripapillary retinal nerve fiber layer and choroidal thickness in eyes with tilted optic disc. J Glaucoma. 2015;24(1):45-50.
PubMedArticle
22.
Pinto  LM, Costa  EF, Melo  LA  Jr,  et al.  Structure-function correlations in glaucoma using matrix and standard automated perimetry versus time-domain and spectral-domain OCT devices. Invest Ophthalmol Vis Sci. 2014;55(5):3074-3080.
PubMedArticle
23.
Nukada  M, Hangai  M, Mori  S,  et al.  Detection of localized retinal nerve fiber layer defects in glaucoma using enhanced spectral-domain optical coherence tomography. Ophthalmology. 2011;118(6):1038-1048.
PubMedArticle
24.
Nukada  M, Hangai  M, Mori  S,  et al.  Imaging of localized retinal nerve fiber layer defects in preperimetric glaucoma using spectral-domain optical coherence tomography. J Glaucoma. 2014;23(3):150-159.
PubMedArticle
25.
Shin  HJ, Cho  BJ.  Comparison of retinal nerve fiber layer thickness between Stratus and Spectralis OCT. Korean J Ophthalmol. 2011;25(3):166-173.
PubMedArticle
26.
Knighton  RW, Qian  C.  An optical model of the human retinal nerve fiber layer: implications of directional reflectance for variability of clinical measurements. J Glaucoma. 2000;9(1):56-62.
PubMedArticle
27.
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(4):266-273.
PubMedArticle
28.
Budenz  DL, Fredette  MJ, Feuer  WJ, Anderson  DR.  Reproducibility of peripapillary retinal nerve fiber thickness measurements with stratus OCT in glaucomatous eyes. Ophthalmology. 2008;115(4):661-666.e4, e664.
PubMedArticle
29.
Leung  CK, Mohamed  S, Leung  KS,  et al.  Retinal nerve fiber layer measurements in myopia: an optical coherence tomography study. Invest Ophthalmol Vis Sci. 2006;47(12):5171-5176.
PubMedArticle
30.
Hwang  YH, Lee  JY, Kim  YY.  The effect of head tilt on the measurements of retinal nerve fibre layer and macular thickness by spectral-domain optical coherence tomography. Br J Ophthalmol. 2011;95(11):1547-1551.
PubMedArticle
31.
Kim  JS, Ishikawa  H, Gabriele  ML,  et al.  Retinal nerve fiber layer thickness measurement comparability between time domain optical coherence tomography (OCT) and spectral domain OCT. Invest Ophthalmol Vis Sci. 2010;51(2):896-902.
PubMedArticle
×