Reproducibility of Anterior Chamber Angle Analyses With the Swept-Source Optical Coherence Tomography in Young, Healthy Caucasians

Abstract

Purpose.: To assess the reproducibility of iridocorneal angle (ICA) analysis in young, healthy Caucasian subjects using swept-source optical coherence tomography (SS-OCT) by determining variability and interobserver agreement between expert and nonexpert observers.

Methods.: Thirty-one healthy volunteers (nonexperts) acquired three consecutive SS-OCT images of the right eyes of their peer nonexperts. Images were analyzed by 31 nonexperts and additionally by three experts, whereby the angle opening distance (AOD) and the trabecular iris space area (TISA) at 500 and 750 μm were calculated. A random intercept model was used to determine the amount of variation between observers. In addition, the intra-observer variability between nonexperts and experts was calculated by determining the coefficient of variation (CV).

Results.: A significant difference was found in the expert analysis for the nasal and temporal angle in the AOD500 (P = 0.002), AOD750 (P < 0.01), and TISA750 (P < 0.01), and the values AOD500 (P = 0.025), AOD750 (P = 0.012), and TISA500 (P = 0.010) were significantly larger if nonexperts analyzed SS-OCT images. The CV was only significant larger for nonexperts for AOD500 (11.1% vs. 8.7%, P < 0.01).

Conclusions.: This study demonstrated high reproducibility of angle analysis in young, healthy Caucasian subjects using SS-OCT. Nevertheless, nonexperts obtained significant larger values compared with experts, implying that training is a necessary requirement before analyzing SS-OCT images in ophthalmic practice.

Introduction

Assessment of the anterior chamber angle is of major importance in diagnosing angle closure. The “gold standard” to diagnose angle closure is gonioscopy. However, the diagnostic value of gonioscopy is limited. It is a subjective assessment, which requires experience for correct interpretation, and substantial interobserver variability has been shown. ¹  

Recently, swept-source optical coherence tomography (SS-OCT; CASIA SS-1000; Tomey Corporation, Nagoya, Japan) has become available. This novel noncontact Fourier-domain OCT is equipped with a swept laser source at a wavelength of 1310 nm to image the anterior segment of the eye, ^2–4 which is capable of precisely delineating fine-angle structures of the anterior chamber angle. ^2,5 Compared with other anterior segment OCT devices (such as AS-OCT, Visante; Carl Zeiss Meditec, Inc., Dublin, CA, USA), the SS-OCT technology promises a superior assessment of the anterior chamber angle (i.e., more accurate visualization of the scleral spur and angle recess), because of the higher resolution of the images (≤10-μm axially and ≤30-μm laterally), higher scanning speed (2,000 A-scans/s vs. 30,000 A-scans/s) and fewer motion artifacts. ^4,5 Another strength of the SS-OCT is that it is capable of acquiring 128 images in a single measurement, allowing anterior chamber angle measurements in any direction, while the AS-OCT can only acquire a single image in a single direction. An advantage SS-OCT has over gonioscopy is its ability to perform an exam under controlled light conditions, without alteration of the anterior chamber angle by inadvertent contact to the eye. However, indentation gonioscopy cannot be replaced by SS-OCT. 

The scleral spur is used as a reference point for the relative position of the trabecular meshwork, from which anterior chamber angle parameters can be calculated to assess the anterior chamber angle. Thus, accurate scleral spur marking is crucial for correct analysis of an OCT-image. 

In the ideal world, a nonexpert (someone with limited knowledge of ophthalmology) should be able to obtain reliable analyses and get reliable estimates of anterior chamber angle parameters with the SS-OCT. Since we have no evidence on the quality of the analyses of nonexperts, we were interested to compare outcomes of experts with nonexperts. The purpose of the present study was therefore to assess the variability of anterior chamber angle values with SS-OCT together with interobserver agreement, for expert and nonexpert observers. 

Methods

Subjects

This prospective study was conducted in accordance with the tenets of the Declaration of Helsinki and was approved by the Maastricht University Medical Center (Maastricht, The Netherlands) institutional review board. 

Healthy volunteers, medical students with a basic knowledge of ophthalmology (nonexperts), were recruited from the University Eye Clinic Maastricht to respectively perform and undergo SS-OCT imaging. After the nature of the experiment had been explained, all volunteers gave their signed informed consent. A subject was considered healthy if he or she was without history of ocular surgery, did not have evidence of ocular disease and did not use topical medication that could affect pupil size. 

The experts were persons working at the University Eye Clinic Maastricht with extensive operational knowledge: an ophthalmic technician, a senior researcher and a PhD student in ophthalmology. They perform many image acquisitions as part of their regular work in ophthalmic care, research, and education. 

Anterior Chamber Angle Acquisition and Analysis

All nonexperts received a verbal and manual instruction by one of the three experts in a standardized way on how to acquire and analyze SS-OCT images (Supplementary Appendix S1). Nonexperts were allowed to use this protocol when obtaining and analyzing images. Images were taken under room light conditions (light intensity >200 lux), with an undilated pupil and subjects looking at the internal fixation light. Nonexperts acquired images in the angle analysis mode of the right eyes of fellow nonexperts. This scan mode consists of 128 radial B-scans, 16-mm long, automatically aligned on the corneal center. Each B-scan includes 512 A-scans. The measurement time with this scan mode is 2.4 seconds. 

The angle parameters were analyzed using the available software for SS-OCT (Version 6; Tomey, Nagoya, Japan). Image analysis was performed by manual placement of the iridocorneal angle (ICA) tool at the scleral spur in the temporal and nasal angle (0°–180° meridian), after which the intrinsic software of the SS-OCT automatically calculated the angle opening distance (AOD) and trabecular iris space area (TISA) at 500 and 750 μm. 

Both the nonexperts as well as the three experts analyzed all nasal and temporal angles of the SS-OCT images after marking the scleral spur according to the protocol (Fig. 1A, Supplementary Appendix S1). To reduce the possible memory bias of the observer regarding the marking of the scleral spur, a time interval of 2 minutes between each analysis was used and images were analyzed in random order. 

Figure 1

 

(A) Location of the scleral spur in a SS-OCT image. (B) The ICA tool is shown, which had to be placed manually on the scleral spur by the operator and the green line measuring the AOD750 (mm). (C) The ICA tool with in the green surface measuring the TISA750 (mm²).

Variability Due to Repeated SS-OCT Image Acquisition by a Single Operator

Three images were taken by one of the fellow, nonexpert peers. A random intercept model was used to assess if the repeated acquisitions were different. The subjects were considered random effects (n = 31), whereas the experts (n = 3) and the repeated image acquisitions (n = 3) were fixed-effect factors. 

Variability Analysis

All data were analyzed using a statistical software package (SPSS 20; SPSS, Inc., Chicago, IL, USA). All available data were extracted from the analyses taken by the experts as well as the nonexperts. First, histograms and frequency analysis were performed to obtain information on distribution of ICA data. Descriptive statistical results were described as the mean ± SD. Next, the difference in mean AOD and TISA between experts and nonexperts was measured. 

In the expert analysis, we determined the interobserver variability for each parameter by using a random intercept model, in which the subjects were considered random effects (n = 31), whereas the experts (n = 3) and the angle (nasal or temporal) were fixed-effect factors, to determine if the expert observers were significant sources of variation in the analysis of the anterior chamber angle parameters. 

To assess the measurement error, the within-subject SD was calculated for the three images per examiner per angle. The intra-observer variability of analyzing a SS-OCT image was determined by obtaining the coefficient of variation (CV) as a measure of reproducibility. The CV was defined as the SD divided by the mean, after which it was multiplied by 100, to express it as a percentage. The CV was calculated separately for the nonexperts and the experts. 

The mean values as well as CV of all four parameters were plotted, with the 95% limits of agreement (LoA) to assess the agreement between nonexperts versus experts. 

A P value less than 0.05 was considered statistically significant. 

Results

Thirty-one eyes of 31 Caucasian subjects, 13 men and 18 women, with a mean age of 24 ± 2 years, were enrolled. The mean spherical equivalent was −1.6 ± 2.5 diopters (D; range, −8.9 to +2.5 D). The 31 nonexperts as well as the three experts analyzed all 93 (31 × 3) images. The scleral spur was identifiable in all SS-OCT images. No images had to be excluded for the reason of inadequate image quality at the angle location. In total, 93 SS-OCT images were used for marking the scleral spur. Due to the inability of the software to identify the correct trace line (demarcation line for the iris and the cornea) in some images, not all anterior chamber angle parameters could be quantified. The percentage for the unidentified trace lines for each parameter were: AOD500 2% to 9%; AOD750 2% to 5%; TISA500 14% to 25%; TISA750 15% to 25%. There was no significant difference between the nasal and temporal angle and between nonexperts versus experts in number of parameters that could not be quantified. However, it should be noted that there was a significant difference between unidentifiable AOD and TISA parameters (P < 0.05). 

Variability Due to Repeated SS-OCT Image Acquisition by a Single Operator

The intercept model used showed that the repeated image acquisitions did not significantly differ (P > 0.4). This was the case for all outcome parameters. 

Variability Analysis

The interobserver variability of the three experts did not differ between the experts in the analysis of the anterior chamber parameters. Thus, the experts were no source of variability in outcome parameters (Table 1). 

Table 1

 

Interobserver Variability Between the Experts Displayed by the SD of the Four Outcome Parameters (Nasal and Temporal Angle) and the Level of Significance Between the Experts

	Expert 1	Expert 2	Expert 3	P Value
SD AOD500	0.058	0.052	0.068	0.07
SD AOD750	0.070	0.069	0.072	0.95
SD TISA500	0.024	0.024	0.028	0.39
SD TISA750	0.042	0.039	0.047	0.39

 Since there was no statistically significant difference in AOD (mm) and TISA (mm²) analyses between the different experts, the values were averaged for the experts per angle in the next tables.

Nasal and Temporal Angle Measurement Comparison

Since there was no statistically significant difference in AOD and TISA analyses between the different experts, the values were averaged for the experts per angle. Table 2 displays the absolute means for the parameters acquired by the three experts in the nasal and temporal angle. The temporal angle showed statistically significant larger parameters than the nasal angle for the AOD500 (P < 0.01), AOD750 (P < 0.01), and TISA750 (P < 0.01). 

Table 2

 

Mean AOD (mm) and TISA (mm²) Analyses Acquired by the Three Experts in the Nasal and Temporal Angle and the Variability Between the Angles Displayed as a Level of Significance

	Nasal Angle	Temporal Angle	P Value
Mean AOD500, mm	0.671	0.707	0.002
Mean AOD750, mm	0.875	0.988	<0.001
Mean TISA500, mm2	0.273	0.275	0.680
Mean TISA750, mm2	0.476	0.502	0.006

Anterior Chamber Angle Analyses Between Nonexperts and Experts

The mean anterior chamber angle parameters as obtained by experts against nonexperts were analyzed. Figure 2 displays Bland-Altman plots of the mean of the four outcome parameters AOD500, AOD750, TISA500, and TISA750 for the nasal and temporal angle. The mean differences for all four parameters were each greater than 0, which represents a difference in absolute values AOD500, AOD750, TISA500, and TISA750 between the nonexperts and experts. Since there was a significant difference between the nasal and temporal angle, a random intercept model was used, in which the subjects were a random effect, whereas the experts/nonexperts and the nasal and temporal angles were covariates. There was a significant difference between the absolute values of AOD and TISA analyses between nonexperts and experts, whereby nonexperts obtained statistically significant larger AOD500 (P = 0.03), AOD750 (P = 0.01), and TISA500 (P = 0.01). Mean differences with 95% confidence interval (CI) are displayed in Table 3 and confirm the results of Figure 2. 

Figure 2

 

Bland-Altman plots showing mean AOD (mm) and TISA (mm²) of both nonexperts and experts and the difference in AOD (mm) and TISA (mm²) between nonexperts and experts in the nasal and temporal angle of all four anterior chamber angle parameters, namely AOD500 (A), AOD750 (B), TISA500 (C), and TISA750 (D). The y-axis represents the difference between nonexperts and experts with the lines indicating the 95% LoA.

Table 3

 

Difference (Δ) in Mean AOD (mm) and TISA (mm²) Analyses Between Nonexperts and Experts With 95% CI and the Level of Significance (P Value)

	Δ	95% CI	P Value
AOD500, mm	0.027	(0.003–0.051)	0.025
AOD750, mm	0.043	(0.009–0.076)	0.012
TISA500, mm2	0.055	(0.013–0.097)	0.010
TISA750, mm2	0.015	(−0.003–0.033)	0.111

 The values were averaged for the experts per angle.

Coefficient of Variation Between Nonexperts and Experts

Figure 3 displays Bland-Altman plots of the CV for all four parameters for both angles. The mean differences in the four outcome parameters was greater than or equal to 0, representing a small difference in CV between the nonexperts and experts. 

Figure 3

 

Bland-Altman plots showing CV of AOD (mm) and TISA (mm²) of both nonexperts and experts and the difference in CV in AOD (mm) and TISA (mm²) between nonexperts and experts in the nasal and temporal angle of all four anterior chamber angle parameters, namely AOD500 (A), AOD750 (B), TISA500 (C), and TISA750 (D). The y-axis represents the difference between nonexperts and experts with the lines indicating the 95% LoA.

Table 4 displays the mean CV (%) of AOD (mm) and TISA (mm²) analyses of the experts and the nonexperts. There was no statistically significant difference between the CV for the nasal or temporal angles. Therefore, the CV was averaged for these positions. The difference between the AOD500 analyses (2.36%; P = 0.005) was statistically significant. There was no statistically significant difference for the outcome parameters AOD750 (0.88%; P = 0.28), TISA500 (0.41%; P = 0.67), and TISA750 (0.33%; P = 0.74). 

Table 4

 

Coefficient of Variation (CV [%]) Of AOD (mm) and TISA (mm²) Analyses of the Experts and the Nonexperts

	Expert, %	Nonexpert, %	P Value
AOD500, mm	8.74	11.1	0.005
AOD750, mm	7.43	8.31	0.275
TISA500, mm2	9.45	9.86	0.665
TISA750, mm2	8.69	9.02	0.740

 CV was defined as the SD divided by the mean, after which it was multiplied by 100, to express it as a percentage. The CV was calculated for the nonexperts and the whole group of experts.

Discussion

To our knowledge, this is the first study comparing the variability in anterior chamber angle parameters of young, healthy eyes, analyzed with the SS-OCT, between experts and nonexperts. A high reproducibility of angle parameters and low CV was found. 

Ideally, to answer the question of reproducibility of angle analysis, we should have analyzed the same image three times. However, we found that the repeated image acquisitions were not different and did not influence the outcome parameters. We, thus, considered our method accurate to study the reproducibility of angle analysis with SS-OCT. 

A recent study comparing the angle landmarks obtained by AS-OCT and SS-OCT showed a visualization of the scleral spur between 70% and 78.9% for AS-OCT ^6,7 and between 95% and 100% for SS-OCT. ⁸ Although the scleral spur could be visualized in all images in our study, it was not always possible to quantify all anterior chamber angle parameters due to inability of the software to identify the correct trace line in a significant number of images. In these images in which the trace line was not correctly identified, it could have been manually manipulated. However, since our goal was to assess the reproducibility and variability of the SS-OCT analyses, the only observer input was to determine the location of the scleral spur in the nasal and temporal angles, without changing the trace line. After obtaining all ICA data, two experts (JB, HR) concluded that changing the trace line is a simple, logical procedure. Nevertheless, this need to sometimes adjust the trace line indicates that SS-OCT images cannot be analyzed without knowledge of the anterior chamber angle anatomy. 

A limitation of our study is that SS-OCT images were assessed in the 0° to 180° meridian only. We did not assess other meridians, although the SS-OCT acquires 128 images in one single measurement and this was possible. Kim et al. ⁹ concluded that the temporal angle was the largest and the inferior angle the smallest. Our analyses confirm the significant difference between nasal and temporal angle values. 

The aim of our study was to address the reproducibility of the SS-OCT as a possible basis for further research with this device. Therefore, initially we chose young, healthy volunteers to study this. We did not assess patients with ocular disease or angle closure, and therefore we could not analyze the reproducibility of angle closure detection with SS-OCT. However, there is already some evidence regarding the angle closure detection capability of AS-OCT. ¹⁰ Presuming SS-OCT, because of the mentioned advantages over AS-OCT, might equal AS-OCT or be superior to it, the angle closure detection should also be equal or superior to AS-OCT. Because this is hypothesized, there is an obvious need for evidence regarding this subject. 

Although Liu and colleagues ¹¹ demonstrated low variability of SS-OCT between experts for analyzing the anterior chamber angle, our study showed larger anterior chamber angle values for nonexperts. The device has minimized several factors that may influence the quality of the image acquisition, however, correctly locating the scleral spur remains an observer input, possibly introducing measurement errors. In accordance with our results, it was confirmed that AS-OCT images can be made by less experienced technicians. ^12,13 However, Tan et al. ¹² also suggested that their results may have been good because the nonexperts received instructions on how to acquire and analyze an image, which was also the case in this study. Despite this instruction, nonexperts obtained significant larger values indicating that examiner experience is of significant importance for the accuracy of the analyses. Training would therefore be advisable before starting analyzing SS-OCT images in ophthalmic practice. 

Acknowledgments

In remembrance of our co-author Marc Frusch. 

The authors thank Frenne Verbakel for his contribution to our study. 

Disclosure: H.C.S. Römkens, None; H.J.M. Beckers, Alcon (C), Allergan (C), MSD (C), Pfizer (C); M. Frusch, None; T.T.J.M. Berendschot, None; J. de Brabander, None; C.A.B. Webers, Alcon (C), Allergan (C), MSD (C), Pfizer (C)