To investigate the additive role of spectral-domain optical coherence tomography (SDOCT) in the structural diagnosis in glaucoma.
Reliability and validity analysis.
Structural examinations from 109 eyes of 109 healthy individuals and 151 eyes of 151 glaucoma patients with different severities were included. Four structural-diagnostic examination sets were prepared using stereo-optic disc photography (SDP), red-free retinal nerve fiber layer photography (RNFLP), and SDOCT: (1) SDP (S), (2) SDP and SDOCT (SO), (3) SDP and RNFLP (SR), and (4) SDP, RNFLP, and SDOCT (SRO). Five glaucoma specialists were instructed to classify subjects as normal or glaucoma using each of the 4 diagnostic sets in the order S, SO, SR, and SRO, with a 1-month interval. The interobserver agreement was evaluated using kappa (κ) statistics. The additive effect of SDOCT on the diagnostic performance of the specialists was evaluated using the generalized estimating equation.
Five glaucoma specialists showed an excellent level of interobserver agreement on the diagnostic assessments based on the 4 sets. In the comparison of the collective diagnostic performance of the specialists, addition of SDOCT to SDP showed an approximately 2-fold significant increase in the diagnostic accuracy. Adding SDOCT to SDP significantly enhanced the specialists’ structural-diagnostic ability with respect to the moderate glaucoma, though not mild or advanced glaucoma.
SDOCT significantly enhanced the diagnostic accuracy of the glaucoma specialists’ performance, showing its additive diagnostic value in judging glaucomatous structural damage, especially in the moderate stage of glaucoma.
Evaluation of structural (optic nerve head [ONH] and retinal nerve fiber layer [RNFL]) and functional (visual field) components is a mainstay of glaucoma diagnostics. Of the two, evaluation of structural changes is the initial, fundamental step in glaucoma diagnosis. Structural changes serve as the primary sign of glaucoma likelihood; they provide the basis of the initial glaucoma-diagnostic indication as to whether patients will undergo further examination or treatment. Thus, accurate evaluation of glaucomatous structural damage is indispensable for appropriate diagnosis of glaucoma.
Spectral-domain optical coherence tomography (SDOCT), with its outstanding performance in optic disc and RNFL thickness measurement, has opened a new era of objective, quantitative assessment of structural glaucomatous damage. Despite its remarkable glaucoma-diagnostic performances confirmed in a number of studies, SDOCT remains an ancillary glaucoma-diagnostic device rather than a “standalone modality.” Most of the published studies, however, evaluated the mere glaucoma-diagnostic performances of optical coherence tomography (OCT) parameters or compared their diagnostic abilities relative to those of other examinations. Therefore, they do not actually represent how influential OCT is to the support of clinical structural assessment or when used in combination with other structural examinations from clinicians’ standpoints.
In our previous report on an additive diagnostic role of SDOCT for glaucoma specialists, we noted that SDOCT had only a limited additive effect on glaucoma specialists’ decisions when employed with combined stereo disc photography (SDP) and standard automated perimetry (SAP). In daily clinical practice, however, a reliable visual field is not always obtainable, and the result often is difficult to interpret, owing to its subjectivity, high level of fluctuation, or false-negative defects mimicking glaucomatous functional loss. In light of these, the present study aimed to investigate the adjunctive structural-diagnostic role of OCT for glaucoma specialists in basic structural optic disc and/or RNFL examinations, without visual field evaluation.
This study included consecutive open-angle glaucoma patients and healthy individuals who had visited the Glaucoma Clinic of Seoul National University Hospital between January 2014 and March 2015. It was approved by the Institutional Review Board of Seoul National University Hospital and adhered to the tenets of the Declaration of Helsinki.
All of the subjects underwent a complete ophthalmologic examination, including measurement of best-corrected visual acuity, refractive error, axial length, central corneal thickness, and intraocular pressure by Goldmann applanation tonometry. Gonioscopy and dilated fundus examination also were performed. Red-free retinal nerve fiber layer photography (RNFLP; Vx-10; Kowa Optimed, Inc, Tokyo, Japan), SDP, and SDOCT (Cirrus-HD; Carl Zeiss Meditec, Inc, Dublin, California, USA) images were obtained for all of the subjects, and SAP with the Swedish interactive thresholding algorithm and 30-2 standard program (Humphrey Field Analyzer II; Carl Zeiss Meditec, Inc) was performed.
Eyes were included if they showed a best-corrected visual acuity of 20/40 or better (regardless of refractive error or axial length) and an open anterior chamber angle on gonioscopic and slit-lamp examination. Eyes were excluded if there was a history of any of the following: intraocular disease or optic neuropathy other than glaucoma, disease that could affect a visual field test, or intraocular surgery other than simple cataract surgery. In cases where both eyes were eligible, 1 eye was randomly selected.
Spectral-domain Optical Coherence Tomography Imaging
The optic disc and RNFL were imaged using Cirrus-HD OCT (software version 6.0), which procedure has been described previously. All of the SDOCT images were acquired by a single, well-trained technician and showed signal strengths of 8 or more, no motion artifacts, no overt misalignment, and no overt decentration of the measurement circle around the optic disc. For the present study, RNFL and ONH algorithms based on the data obtained from a 200 × 200 cube optic disc scan were employed. The ONH and RNFL analysis printout was used as an SDOCT image containing an RNFL thickness map, an RNFL deviation map, a temporal-superior-nasal-inferior-temporal RNFL thickness plot, quadrant and clock-hour maps with RNFL thickness measurements and diagnostic classification, and ONH-parameter analyses.
Classification of Diagnoses
A total of 109 healthy individuals and 151 patients who had been diagnosed with and followed up on for glaucoma at least 6 months before the enrollment period were included in the present study. Two experienced glaucoma specialists (K.H.P. and K.E.K.) reviewed each subject’s examination and clinical information in an independent manner, classifying eyes as either “glaucoma” or “healthy individual.” In cases of disagreement, the final decision was made by a third observer (D.M.K.).
The glaucoma diagnosis was based on a combination of structural (SDP and red-free RNFLP) and functional (SAP) measurements. The present study included eyes with perimetric glaucoma having a glaucomatous optic disc appearance and corresponding glaucomatous visual field defect. Glaucomatous optic neuropathy was defined as characteristic optic disc changes (eg, rim notching, thinning, increased cup-to-disc ratio, and/or RNFL abnormalities). Glaucomatous visual field defect was defined as (1) the presence of a cluster of 3 or more non-edge points within P < .05 of being normal, 1 with P < .01 on pattern deviation plot; (2) a pattern standard deviation with P < .05; or (3) a glaucoma hemifield test result outside the normal limits, when proved by 2 consecutive tests within a month. Only reliable fields with false-positive and false-negative error rates of 15% and fixation loss of 15% or less were included. Glaucomatous eyes were further divided into groups of different severity based on the SAP mean deviation (MD): early (MD >−6 dB), moderate (−12 < MD ≤ −6 dB), and advanced (MD ≤−12 dB) stages.
Healthy eyes were those of individuals from the general population who had visited the clinic for regular health check-ups, correction of refractive error, or dry-eye treatment. Those with no history of IOP >21 mm Hg, no history of IOP-lowering medication use, no family history of glaucoma, no evidence of ocular disease, no abnormal appearance of optic disc (eg, absence of glaucomatous optic neuropathy, pallor, swelling) or RNFL, and a normal visual field were included.
Preparation and Evaluation of Diagnostic Combination Sets of Structural Examinations
To examine the additive diagnostic role of SDOCT in structural diagnosis, sets containing different combinations of structural examinations were prepared and sent to glaucoma specialists for evaluation. Four combination sets of structural examinations were prepared for each subject using SDP, RNFLP, and SDOCT: (1) set S: SDP, (2) set SO: SDP and SDOCT, (3) set SR: SDP and RNFLP, and (4) set SRO: SDP, RNFLP, and SDOCT. Because optic disc examination is considered to be an initial step in glaucoma diagnosis, combination sets were created by adding RNFLP, SDOCT, or both to the S set. Each set was composed of the single most recent reliable image from each device. Images that were taken on the same day were used as much as possible, but in cases where this was not feasible, a 4 month interval between examinations was allowed. Additionally, the same image from each device was included in each set: the same SDP image was used for the S, SO, SR, and SRO sets; the same SDOCT image for the SO and SRO sets; and the same RNFLP image for the SR and SRO sets. Each subject’s images were saved as JPEG files, and only the relevant images were put into the specific folder for the corresponding combination set: folders in set S contained only the SDP images, those in set SO contained only the SDP and SDOCT images, and so on. In each combination set, the folders containing the subjects’ examinations were randomly sorted and numbered independently of diagnosis or of glaucoma severity. Additionally, when each combination set was created, the folders were newly assigned in random order.
These combination sets were sent to 5 fellowship-trained glaucoma specialists (S.H.K., J.W.J., M.H.S., J.H.S., M.K.) from different professional medical institutes, who had been invited and signed informed consents to take part in this study. They were routine users of the provided examinations, including Cirrus-HD OCT. All of the glaucoma specialists received the combination sets in the order of sets S, SO, SR, and SRO. They were allowed to evaluate each set within a week, and a 1-month interval between set evaluations was provided. For each set, they were instructed to evaluate eyes as normal or glaucoma in an independent manner, being masked to all of the relevant information and to any of the previous readings. They were allowed to diagnose without specified guidelines, based on their knowledge and clinical experience alone.
For the comparison of group characteristics, 1-way analysis of variance with post hoc Tukey test was used for continuous variables and χ 2 test or Fisher exact test for categorical variables. Kappa (κ) statistics were used to evaluate the interobserver agreement among the glaucoma specialists’ glaucoma diagnoses. The strength of agreement was categorized according to the method of Landis and Koch : 0 = poor, 0 to 0.20 = slight, 0.21 to 0.40 = fair, 0.41 to 0.60 = moderate, 0.61 to 0.80 = substantial, and 0.81 to 1.00 = almost perfect. The McNemar test was used to compare individual diagnostic sensitivity and specificity. To compare the collective diagnostic performance of the glaucoma specialists according to the different diagnostic combination sets and the differing glaucoma severity, the generalized estimating equation (GEE), a method for estimating generalized linear model parameters with a possible unknown correlation between outcomes, was used. The GEE was also used for quantitative evaluation of how each combination set increased the sensitivity and specificity of the specialists’ assessments, presenting as an odds ratio (OR) with 95% confidence interval (CI). To adjust for multiple comparisons, Bonferroni correction was applied.
A total of 109 healthy individuals and 151 glaucoma patients with different glaucoma severities were included in the present study. The mean age of the included subjects was 59.0 ± 10.4 years, and the mean spherical equivalent was −2.97 ± 3.17 diopters. No significant difference was found in the age, sex, or refractive error between the healthy individuals and glaucoma groups ( Table 1 ). However, in the patients with more advanced stages of glaucoma, significantly larger average cup-to-disc ratio, thinner RNFL, and worse visual field MD and pattern standard deviation value s were noted.
|Control (N = 109)||Glaucoma (N = 151)||P|
|Early (N = 52)||Moderate (N = 50)||Advanced (N = 49)|
|Age, y||58.6 ± 8.2||57.6 ± 9.4||58.3 ± 12.9||62.4 ± 12.7||.086 a|
|Male, n, (%)||58 (53.2)||36 (69.2)||22 (44.0)||30 (61.2)||.057 b|
|SE, diopters||−3.12 ± 3.19||−2.76 ± 2.87||−2.93 ± 3.31||−2.34 ± 3.66||.669 a|
|Axial length, mm||24.7 ± 1.4||25.1 ± 1.4||24.6 ± 1.6||24.6 ± 1.5||.306 a|
|CCT, μm||550.5 ± 31.0||545.3 ± 32.3||537.3 ± 23.7||536.9 ± 39.1||.073 a|
|Disc area, mm 2||1.96 ± 0.49||1.79 ± 0.47||1.94 ± 0.48||1.95 ± 0.43||.162 a|
|Average cup-to-disc ratio||0.62 ± 0.15||0.69 ± 0.11||0.74 ± 0.09||0.82 ± 0.08||<.001* a|
|Average RNFL thickness, μm||91.8 ± 9.0||76.9 ± 9.8||70.2 ± 9.6||60.7 ± 8.5||<.001* a|
|MD, dB||0.06 ± 1.63||−2.61 ± 1.57||−8.56 ± 1.87||−16.24 ± 3.45||<.001* a|
|PSD, dB||1.84 ± 0.59||5.11 ± 2.78||11.81 ± 3.41||13.86 ± 2.27||<.001* a|
a One way analysis of variance.
A very high level of interobserver agreement was observed among the specialists’ glaucoma diagnoses: almost perfect levels of agreement for the S (κ = 0.829), SO (κ = 0.906), SR (κ = 0.907), and SRO (κ = 0.909) sets ( Table 2 ). Moreover, the agreement increased with increasing number of diagnostic examinations.
|Combination Sets of Structural Examinations||Interobserver Agreement (κ)|
|SDP + SDOCT||0.906 (0.866–0.937)|
|SDP + RNFLP||0.907 (0.868–0.939)|
|SDP + RNFLP + SDOCT||0.909 (0.871–0.939)|
In the individual performance analysis ( Table 3 ), SDOCT increased the glaucoma-diagnostic sensitivity when added to SDP for all of the glaucoma specialists, though not significantly after application of Bonferroni correction. For 2 glaucoma specialists, the addition of RNFLP to SDP significantly increased the sensitivity. The diagnostic performances of the specialists’ assessments using SO and SR were similar. The sensitivity of the diagnostic assessments also showed no difference when SDOCT was added to SR.
|Glaucoma Specialists||Diagnostic Combination Sets||Comparison of Diagnostic Performance Between Sets|
|S||SO||SR||SRO||S vs SO||S vs SR||SO vs SR||SO vs SRO||SR vs SRO|
|Sensitivity, %||P a|
|Specificity, %||P a|
In contrast to the individual outcomes, the collective sensitivity of the 5 glaucoma specialists showed a 2-fold increase when using SO relative to S (OR = 2.21, 95% CI 1.36–3.59, P = .001; sensitivity difference 4.0%, Table 4 ). The diagnostic sensitivity using SR showed a larger increment, about a 6-fold increase, relative to S (OR = 5.82, 95% CI 2.71–12.49, P < .001; sensitivity difference 6.2%). As expected, with the addition of both OCT and RNFLP, the diagnostic performance was enhanced approximately 7 times compared with SDP only (OR = 6.51, 95% CI 3.50–12.13, P < .001; sensitivity difference 6.4%). Meanwhile, no significant diagnostic performance difference was found between the SO and SR sets. The addition of RNFLP to the SO set significantly improved the diagnostic ability, whereas no significant change was found the other way around.
|Diagnostic Combination Sets||Sensitivity||Specificity|
|Odds Ratio (95% CI)||P||Odds Ratio (95% CI)||P|
|S vs SO||2.21 (1.36–3.59)||.001*||0.69 (0.37–1.29)||.247|
|S vs SR||5.82 (2.71–12.49)||<.001*||1.18 (0.68–2.02)||.563|
|S vs SRO||6.51 (3.50–12.13)||<.001*||1.15 (0.66–2.00)||.615|
|SR vs SRO||1.12 (0.55–2.29)||.759||0.98 (0.58–1.64)||.935|
|SO vs SR||2.64 (1.02–6.79)||.045||1.70 (0.88–3.29)||.118|
|SO vs SRO||2.95 (1.44–6.04)||.003*||1.66 (1.07–2.57)||.024|