Agreement among three types of spectral-domain optical coherent tomography instruments in measuring parapapillary retinal nerve fibre layer thickness
- Department of Surgery, Division of Ophthalmology, Kobe University Graduate School of Medicine, Kobe, Japan
- Correspondence to Dr Akiyasu Kanamori, Kobe University Graduate School of Medicine, Kobe, 7-5-1, Kusunoki-cho, Chuo-ku, Kobe 650-0017, Japan;
Contributors AK designed data collection tools, monitored data collection for the whole trial, wrote the statistical analysis plan, cleaned and analysed the data, and drafted and revised the paper. MN analysed the data, and drafted and revised the paper. MT and YY designed data collection tools and monitored data collection for the whole trial. YY monitored data collection and revised the draft paper. AN revised the draft paper.
- Accepted 18 January 2012
- Published Online First 14 February 2012
Backgrounds/aims To evaluate the agreement of parapapillary retinal nerve fibre layer (RNFL) thickness among three spectral-domain optical coherence tomography (OCT) instruments.
Methods Two hundred and three glaucomatous eyes and 88 normal eyes were imaged by Cirrus, RTVue and 3D OCT. The average and the four quadrant RNFL thicknesses were evaluated. Agreement among RNFL measurements was evaluated using Bland–Altman analysis and linear regression analysis. The percentage of each quadrant in the average RNFL thickness value was compared among the three instruments.
Results Cirrus showed significantly smaller thickness values than RTVue (difference=8.8 μm, p<0.0001) and 3D OCT (difference=8.1 μm, p<0.0001). Although RNFL measurements among the instruments were highly correlated, the Bland–Altman analysis revealed proportional biases for most of the pair-wise agreements. Additionally, 3D OCT showed strong proportional biases with RTVue and 3D OCT. RTVue had a smaller occupied proportion of nasal quadrants (30.2%) and a larger proportion of inferior quadrants (32.4%) compared with Cirrus and 3D OCT.
Conclusions RNFL measurements among the instruments were well correlated but had different values for thickness. The measurement circle of RTVue might be more superior-temporally located compared with the other instruments. Differences in the measurement protocols might be affected by the disagreements. These instruments should not be used interchangeably.
- Spectral-domain optical coherence tomography
- retinal nerve fibre layer thickness
- Bland–Altman analysis
- experimental and animal models
- intraocular pressure
- diagnostic tests/investigation
- experimental and laboratory
Glaucoma is an optic neuropathy characterised by a specific and progressive injury to the optic nerve and retinal nerve fibre layer (RNFL).1 Evaluation of RNFL plays an important role in the diagnosis and management of glaucomatous patients. Several studies have demonstrated that an early version of optical coherence tomography (OCT), that is, time-domain (TD) OCT, could objectively detect a specific pattern of reduction in the average or focal RNFL thickness. The recently developed spectral-domain (SD) OCT has an enhanced spatial resolution and shortened acquisition time, resulting in an increased reproducibility compared with TD-OCT. However, several independent companies have launched SD-OCT devices of their own with different acquisition speeds and resolution under modified algorithms and principles. Nevertheless, there is only one report that has tested the agreement among three SD-OCT instruments in measuring the RNFL thickness in a single subject population, although many studies have compared the diagnostic ability of SD-OCT with TD-OCT for glaucoma.2–6 In a separate report, Spectralis, Cirrus and RTVue were compared and different values in the parapapillary RNFL thickness were seen; 7 however, these instruments showed similar diagnostic abilities with regard to glaucoma in the same population.8
The purpose of this study was to assess the agreement in measuring the parapapillary RNFL thickness evaluated by Cirrus, RTVue and 3D OCT in normal and glaucomatous individuals under 60 years of age to minimise the ageing effect on the RNFL thickness measurement.
Materials and methods
In this observational cross-sectional study, subjects were recruited from Kobe University Hospital (Kobe, Japan). The study protocol was approved by the Institutional Review Board of Kobe University and adhered to the tenets of the Declaration of Helsinki. Written informed consent was obtained from each subject after an explanation of the study protocol.
All subjects received a full ocular examination. A visual field (VF) test was performed by a Humphrey Field Analyzer 30-2 SITA standard programme (HFA, Carl Zeiss Meditec, Inc., Dublin, CA). To be included, subjects between 20 and 60 years of age had to have a best corrected visual acuity of 20/40 or better, spherical refraction ±8.0 D, cylinder correction within ±3.0 D and open angles. Axial length was acquired with IOL Master (Carl Zeiss Meditec, Inc.). No subjects had undergone any ocular surgeries. VF tests and measurements of the parapapillary RNFL thickness using three SD-OCT instruments were obtained during the next 6 months.
Glaucomatous eyes had open angle and a VF loss corresponding to glaucomatous optic neuropathy. This was defined as vertical cup-disc asymmetry between fellow eyes of 0.2 or more and neuroretinal rim damages such as excavation, rim thinning, and notches with or without parapapillary haemorrhages or RNFL defects. The evaluation of glaucomatous VF defects was based on liberal criteria (two or more contiguous points with a pattern deviation sensitivity loss of p<0.01, or three or more contiguous points with sensitivity loss of p<0.05 in the superior or inferior arcuate areas, or a 10 dB difference across the nasal horizontal midline at two or more adjacent locations and an abnormal result in the glaucoma hemifield test).9 Eyes with any vitreoretinal diseases were excluded. Healthy subjects at least with 20 years of age were recruited as normal controls. Exclusion criteria included the following: intraocular pressure>21 mm Hg, unreliable HFA results (fixation loss or false-positive or false-negative>33%), abnormal findings in HFA suggesting glaucoma according to the Anderson and Patella criteria, any abnormal VF loss consistent with ocular disease, evidence of vitreoretinal diseases and optic nerve or RNFL abnormality.
RNFL thickness measurements
The Cirrus OCT (software V.22.214.171.124, Carl Zeiss Meditec Inc.) used the optic disc cube protocol. This protocol was based on a 3D scan of a 6×6 mm2 area centred on the optic disc. Then, a 3.46 mm diameter circular scan was automatically placed around the optic disc, and the information about parapapillary RNFL thickness was obtained. Images with signal strength <5 were excluded. If the optic disc edge was not located within a scan circle or the centre of the scan circle fell outside the cupping of the disc when the disc was small, the circular scan was manually moved to let the measurements become centred on the optic disc centre. RNFL thickness at 256 measurements points on the circular scan were exported to a personal computer and then evaluated with the RNFL parameters described below.
The RTVue-100 OCT (software V.126.96.36.199, Optovue Inc., Fremont, California, USA) used the optic nerve head map protocol. This protocol generates an RNFL thickness map that was measured along a circle 3.45 mm in diameter centred on the optic disc. The optic nerve head boundary is manually defined within the 3D Disc scan. After the RNFL map was obtained, the RNFL thickness parameters were estimated by assessing a total of 2325 data points between the anterior and posterior RNFL borders. Only good quality images, as defined by a signal strength index of >30, were used. RNFL thickness of each parameter calculated by the original software was used.
The 3D OCT 2000 (software V.7.01, Topcon Inc, Tokyo, Japan) used a 6×6 mm scan protocol centred on the gravity centre of the optic disc. To obtain more accurate circle sizes during the measurement, the magnification effect in each eye was corrected according to the formula provided by the manufacturer (modified Littman's method), which was based on the refraction, corneal radius and axial length. Images with a quality factor of more than 60% were used. If the optic disc edge was misaligned, the circular scan was manually moved as described above. RNFL thicknesses at 1024 points on a circle with a 3.46 mm diameter was exported to a personal computer and then evaluated with the RNFL parameters described below.
Mean 360° RNFL thickness was defined as the average of RNFL thickness. When the temporal margin was designated as 0°, the mean quadrant RNFL thickness was defined as temporal (315°–45°), superior (45°–135°), nasal (135°–225°) and inferior (225°–315°).
This study conducted four statistical analyses. First, one-way repeated-measures analysis of variance was used to determine whether RNFL thickness, measured with the three types of instruments, significantly differed. Second, to assess the possibility that the three instruments had different distributions in circular RNFL thickness, percentage of RNFL thickness in each quadrant relative to the overall mean RNFL thickness were calculated in normal eye group. Glaucomatous eyes were excluded from this analysis because the reduction of RNFL thickness differs among the quadrants. The proportion in each quadrant was compared among the three instruments by one-way repeated-measures analysis of variance. Third, the agreements in RNFL measurements by the three instruments were investigated with Bland–Altman plots.10 11 The difference in each pair of instruments was plotted against the mean. To formally evaluate this relationship, the difference between the two instruments was regressed on their average. If the slope of the regression line was statistically significant, we considered the existence of a proportional bias. Finally, to assess the correlation between the three instruments, we calculated the coefficients of determination (R2) for pair-wise measurements.
Statistical analysis was performed using Medcalc (Mariakerke, Belgium). A p value of <0.05 was considered to be statistically significant.
Two hundred and three glaucomatous eyes from 126 subjects and 88 normal eyes from 61 subjects were enrolled in this study. Table 1 shows the demographics and ocular characteristics of the subjects. Age, refraction and axial length did not show any significant differences between glaucomatous and normal subjects.
Table 2 shows the mean values of RNFL thickness parameters measured by each instrument. Regarding the average RNFL thickness, Cirrus showed significantly smaller values than RTVue (difference=8.8 μm, p<0.0001) and 3D OCT (difference=8.1 μm, p<0.0001). There was no difference in average RNFL thickness measurements between RTVue and 3D OCT (difference =0.7 μm, p=0.568). As for quadrant RNFL thickness, Cirrus measurements were statistically smaller in all quadrants compared with 3D OCT measurements and were smaller in all but the nasal quadrant compared with RTVue measurements. RTVue thickness measurements had higher values in the inferior quadrant and lower values in the nasal quadrant compared with the 3D OCT measurements.
Table 3 summarises the occupied proportion of the quadrant to average RNFL thickness values obtained from the three instruments. The quadrant RNFL thickness in either the superior or inferior quadrant occupied approximately 30% of the average RNFL thickness in all three instruments. Temporal and nasal quadrants occupied approximately 20% and 17%, respectively, of the average RNFL thickness. RTVue measurements had a smaller proportion of occupation in the nasal quadrant and a larger proportion in the inferior quadrant compared with the Cirrus and 3D OCT measurements. In contrast, there were no significant differences in the occupied proportion of all quadrants between Cirrus and 3D OCT. Figure 1 shows a schema for the occupied proportions of quadrants and the measurement circle in each instrument.
Table 4 and figure 2 show Bland–Altman plots for the RNFL thickness measured with the three instruments. Agreements between the RTVue and 3D OCT measurements, or between the Cirrus and 3D OCT measurements, had a proportional bias in all RNFL sectors. Additionally, the proportional bias was found to be in agreement between the Cirrus and RTVue measurements in the superior and inferior RNFL quadrants. Figure 2 shows the agreements among the three instruments and the 95% limits of agreement for the average RNFL thickness. The subanalysis with Bland–Altman plots separately performed in the normal eyes and glaucomatous eyes were shown in table 5. The similar proportional bias was found in the glaucomatous eyes, not in the normal eyes.
Table 6 shows the coefficients of determination for all pair-wise comparisons. As shown in figure 3, all correlations had statistically significant differences (p<0.001). The highest correlation (R2=0.917) was seen between the average thickness measurements by Cirrus and RTVue. Overall, RNFL thickness in the nasal quadrant had a weaker correlation than the other quadrants in all instruments.
This study assessed the agreement in measuring RNFL thickness among Cirrus, RTVue and 3D OCT. RNFL thickness by Cirrus was thinner than the other two devices. The 3D OCT measurement had a proportional bias with the RTVue and Cirrus measurements. The occupied proportions in the quadrant sectors showed a discrepancy among the instruments, suggesting a variability of centre alignment for the scanning circle to the optic disc.
This study excluded subjects older than 60 years of age to minimise the effects of ageing and possibly related cataracts on RNFL measurements. Previous studies have shown a significant association between the magnitude of signal strength and the RNFL thickness.12 13 Several factors, which include age, race and cataracts, may influence the performance of SD-OCT, most notably signal strength. Cataracts may independently affect the acquisition of good images with OCT and also decrease RNFL thickness.14 Alternatively, cataracts may be a comorbid condition that reduces the signal strength, given that ageing per se may reduce the RNFL thickness detected using OCT.15 16
Cirrus measurements indicated thinner RNFL thickness than did the RTVue measurements in all but the nasal quadrants. Two reports have demonstrated, using a single population, the agreement among Cirrus, RTVue and Stratus-OCT, all of which concluded that the RNFL thickness measurements using Cirrus measurements were smaller than those of RTVue.17 18 In addition, Leite et al reported the same result in the comparison of Cirrus, RTVue and Spectlaris.7 Our results are consistent with these reports. On the other hand, the values of Cirrus measurements were also consistently lower than those of the 3D OCT. Only one study compared the RNFL thickness measured by Cirrus and 3D OCT, which showed measurements obtained from Cirrus to be thinner than those from 3D OCT in normal eyes.19 As for the comparison between RTVue and 3D OCT, measurements obtained with RTVue were thicker than those from 3D OCT in the inferior quadrant and thinner in the nasal quadrant. To our knowledge, there are no reports that compare the assessment of RNFL thickness by RTVue and 3D OCT.
Each instrument's software is based on the same principal algorithms for distinguishing the RNFL from the retinal ganglion cell layer. Nevertheless, there were considerable variations in measured RNFL thickness among instruments. In addition, there were significant discrepancies in quadrant RNFL thickness measurements among devices, which may be at least in part accounted for by the variation of the occupied proportion. RTVue had a significantly lower occupation proportion in the nasal quadrant and a higher proportion in the inferior quadrant than Cirrus and 3D OCT. Given that the more distal to the optic disc, the thinner the RNFL thickness, the alignment centre of the scan areas of RTVue measurement might be located more superior-temporally, compared with Cirrus and 3D OCT. Figure 1 depicts the presumed spatial correspondence of the circle scans from the three devices on the parapapillary retinal surface. Note that the RTVue applied the circular scan for RNFL measurements, while the other two instruments utilised the square scan. In addition, the edge of the optic disc was manually determined on acquisition using RTVue, whereas the other two instruments automatically defined the optic disc margin. The automatic identification of the optic disc boundary was well performed in most of the cases. However, Cirrus misaligned four eyes and 3D OCT six eyes and manual alignment of the measurement circle was needed in these eyes. These differences in the scanning methods and decision of disc boundary probably affected the acquisition of the circular measurement of RNFL thickness. Therefore, the RNFL thickness measurements are not interchangeable among SD-OCT instruments.
Superior and inferior quadrants were found to be made up of approximately 30% each of the whole circumpapillary RNFL. In contrast, the temporal and nasal quadrants were approximately 20% each. These estimated occupied proportions of the RNFL measured by OCT were consistent with a histological study using human postmortem eyes.20 However, it is hard to identify which data among the three instruments reflect actual profiles of RNFL. Direct comparative studies between OCT measurements and histological analysis are needed to evaluate this issue. However, caution must be taken given the fact that the RNFL thickness by SD-OCT was apparently thicker than the histological sections on the same circumpapillary circle with 3.5 mm diameter from normal human eyes.20 Methodological artefacts may be included in the estimate during the preparation of the histological specimens, such as postmortem tissue swelling and shrinkage.20 21
Although the absolute values of the RNFL thickness varied among the instruments, there were high correlations in all pair-wise comparisons. The average thickness had the strongest correlation as expected. The nasal quadrant had an apparently lower correlation than the other quadrants. Previous studies comparing TD-OCT and SD-OCT also showed a lower correlation and reproducibility in the nasal quadrant.3 22–24 This might be explained by the fact that the incidence angle of the laser beam has a dimmer light in the nasal quadrant.25
The Bland–Altman analysis disclosed that the 3D OCT had a strong proportional bias versus the Cirrus and RTVue. In other words, the thinner the RNFL thickness measurements, the more pronounced change the 3D OCT displayed than RTVue and Cirrus. This finding seems to contradict a previous study that investigated the agreement among Stratus-OCT, Cirrus and 3D OCT-1000.19 However, that previous study evaluated RNFL measurements from normal eyes only. The narrower range of the studied RNFL thickness may make it difficult to estimate the correlation between the average and the difference of two variables. The subanalysis with Bland–Altman plots showed the proportional bias in only glaucomatous eyes (table 5) in this study. In addition, we used 3D OCT-2000 with a recent version (V.7.01) instead of 3D OCT-1000 (V.3.01) that the previous researchers used.19 The reason for the proportional bias of 3D OCT might be related to the correction of ocular magnification, given the fact that axial length and refraction affect RNFL measurements via the ocular magnification.26 27 The Littman correction was performed in only 3D OCT, not Cirrus and RTVue. Absence of the proportional bias between Cirrus and RTVue encouraged this speculation. Further analysis is needed to address the effect of the ocular magnification correction on RNFL measurements by SD-OCT.
In conclusion, although the parapapillary RNFL thickness measurements using the three SD-OCT instruments correlated well, there were considerable differences in the absolute values. The protocol for the scan and analysis software for RNFL measurements might affect the discrepancy. To identify which OCT instrument reflects the most accurate measurement of RNFL thickness, further studies combined with histological methods are needed.
Funding Supported by Grant-in-Aid 22390324 (AN, YY, MN), 20592043 (AK, MN, AN) and 23791983 (AK) for scientific research by the Ministry of Education, Culture, Sports, and Science and Technology of the Japanese government, and Suda Memorial Foundation (AK).
Competing interests None.
Patient consent Obtained.
Ethics approval The study protocol was approved by the Institutional Review Board of Kobe University and adhered to the tenets of the Declaration of Helsinki.
Provenance and peer review Not commissioned; externally peer reviewed.