A Comparison of OCT Parameters in Identifying Glaucoma Damage in Eyes Suspected of Having Glaucoma

Abstract

Purpose:

To compare retinal nerve fiber layer thickness (RNFLT) and Bruch’s membrane opening minimum rim width (BMO-MRW) measured by spectral domain optical coherence tomography (SDOCT) for diagnosing glaucoma in those suspected of having the disease.

Design:

Observational cohort study

Participants:

113 eyes from 81 patients suspected of having glaucoma based on optic nerve appearance.

Methods:

Participants were imaged using SDOCT and RNFLT and BMO-MRW were measured. All participants had normal visual fields at the time of imaging, but were considered suspects based on the appearance of the optic disc during clinical examination. Eyes were classified as glaucomatous or non-glaucomatous based on documented stereophotographic evidence of progressive glaucomatous change in the appearance of the optic nerve or retinal nerve fiber layer occurring before the imaging session. For each parameter, we calculated the area under the receiver operating characteristic (ROC) curve and the sensitivity with fixed specificities of 80% and 95%.

Main Outcome Measures:

Areas under the ROC curves

Results:

Of the 113 eyes suspected of having glaucoma, 52 (46.0%) eyes had progressive glaucomatous optic nerve changes and were classified as having pre-perimetric glaucoma and 61 (54.0%) eyes did not have progressive glaucomatous optic nerve changes (followed untreated for an average of 9.4±3.2 years) and were classified as normal. The areas under the ROC curves were 0.89 (95% CI: 0.84–0.95) for global RNFLT and 0.75 (95% CI: 0.65–0.85) for global BMO-MRW (p=0.006). The sensitivity at 95% specificity was 60% for the global RNFLT parameter and 40% for the global BMO-MRW parameter. The RNFLT parameters that achieved areas under ROC curve ≥0.80 were global (0.89), supero-temporal (0.80), infero-temporal (0.87), and supero-nasal (0.81). The only BMOBRW parameter that achieved area under ROC curve ≥0.80 was infero-temporal (0.82).

Conclusions:

Our findings suggest that RNFLT parameters may be better able to identify pre-perimetric glaucomatous damage in glaucoma suspects than BMO-MRW.

Introduction

Optical coherence tomography (OCT) is a well-established imaging technique commonly used to assist in diagnosing glaucoma, as well as in monitoring patients with the disease.¹ When used as an ancillary diagnostic test in patients suspected of having glaucoma, OCT imaging aims to provide information that can assist clinicians in deciding whether an eye has glaucomatous damage or not, which can then be used for management decisions. To be useful as an ancillary diagnostic test, results from an OCT exam should provide additional information besides what can be promptly gathered from the standard clinical exam, such as ophthalmoscopy, and assessment of visual fields. As a matter of fact, it can be easily argued that a patient presenting with clearly defined glaucomatous visual field loss that is compatible with clinical exam of the optic disc will rarely benefit from an OCT if acquired for the purpose of assisting in the diagnosis, although the test would still most certainly be helpful for longitudinal follow-up and monitoring of progression.

Despite the apparent triviality of the above observations, it is surprising that the vast majority of studies in the literature aiming to investigate the diagnostic accuracy of OCT parameters have included as cases mostly patients with glaucomatous visual field loss and as controls only subjects that show no suspicious signs for the disease.^2–6 Although such studies can be helpful for an initial investigation of the accuracy of these parameters, they do not enroll the clinically relevant population of patients suspected of having the disease. Therefore, it is unlikely that they can helpful in deciding whether OCT provides information to assist diagnosing glaucoma in these cases, or which OCT parameter performs best in this regard.

Two important OCT parameters that can be used to assess structural damage in glaucoma are retinal nerve fiber layer thickness (RNFLT) and measurements of the neuroretinal rim, such as the Bruch’s membrane opening minimum rim width (BMO-MRW). RNFLT measures the thickness of nerve tissue in the peripapillary area and BMO-MRW is a measure of the minimum thickness of nerve tissue at the opening of Bruch’s membrane, averaged around the disc.^7,8 Prior publications have suggested that RNFLT and BMO-MRW measures are equivalent in their ability to identify glaucoma or that BMO-MRW may be slightly better.^7,9–11 However, all of these publications have evaluated clearly glaucomatous populations with well-defined visual field loss compared to healthy controls. It is not known which parameter would perform better at providing diagnostic information that could help diagnose the disease in those suspected of glaucoma, a much more clinically relevant question.

It is challenging to conduct diagnostic studies evaluating glaucoma suspects because there is no independent gold-standard that can be administered at a single point in time to distinguish between suspects who do or do not have glaucoma. However, longitudinal evaluation of optic disc photographs taken at different time points can be used to identify the presence of progressive damage to the optic nerve, which can be used to confirm the diagnosis of disease even in the absence of clearly defined visual field loss.¹² The use of such longitudinal information can serve as reference standard in a study designed to evaluate the diagnostic accuracy of imaging tests when applied in suspects, as reported previously in the literature.^12–14

In the present study, we investigate and compare the accuracy of RNFLT and BMO-MRW in providing diagnostic information in patients suspected of having glaucoma. Long-term follow-up was used to establish diagnosis in these patients and as a reference standard for comparison of accuracies.

Methods

Study population

The patients evaluated in this study were part of a longitudinal cohort study designed to evaluate structural damage and functional impairment in glaucoma. All participants provided informed consent and the Duke University Health System Institutional Review Board approved the study. Data was de-identified and used for statistical analyses. The study adhered to the tenets of the Declaration of Helsinki.

Participants received a comprehensive ophthalmologic exam, which consisted of a review of medical history, evaluation of best-corrected visual acuity (BCVA), slit-lamp biomicroscopy, gonioscopy, IOP measurement using Goldmann applanation tonometry, dilated fundoscopic examination, stereoscopic optic disc photography, and standard automated preimetry using the 24–2 Swedish Interactive Threshold Algorithm (Carl Zeiss Meditec Inc., Dublin, CA). Participants were required to have BCVA of 20/40 or better, open angles on gonioscopy, spherical visual correction within ±5.0 D, cylinder visual correction within ±3.0 D, and no coexisting retinal disease, uveitis, or nonglaucomatous optic neuropathy.

For our study, we identified a subset of participants who were suspected of having glaucoma based on the presence of optic nerve findings suspicious for glaucoma, but who had normal visual fields at the time of the OCT imaging session. Eyes suspected of having glaucoma presented with suspicious neuroretinal rim thinning, cupping, or suspicious/abnormal RNFL defects with or without elevated intraocular pressure. A normal visual field was defined as a mean deviation and pattern standard deviation with P>5% and a glaucoma hemifield test within normal limits.

The eyes suspected of having glaucoma were then classified as glaucomatous or normal based on the presence of progressive glaucomatous change on the sterophotographs of the optic disc taken prior to the imaging session. Two independent, masked graders reviewed the stereoscopic photographs of the optic disc using a stereoscopic viewer. For each eye, the most recent sterophotograph was compared with the oldest available one. Glaucomatous optic nerve change was defined as thinning of the neuroretinal rim, increased excavation, or enlargement of RNFL defects. Changes in rim color, presence of disc hemorrhage, or progressive peripapillary atrophy were not part of the definition of glaucomatous optic nerve change. Differences between the two graders were resolved by consensus or review by a third grader when needed. Eyes without any evidence of progressive glaucomatous change in the appearance of the optic disc were classified as normal and used as the control group. Controls were excluded from our study if they had any history of repeatable visual field defects prior to their imaging session or had received IOP-lowering treatment in the past. A similar study design has been previously used in other studies.^12–14

All patients underwent testing with spectral domain optical coherence tomography (SDOCT) (Spectralis, Heidelberg Engineering Inc., Franklin, MA, USA). Spectralis SD OCT was used to calculate both RNFLT and BMO-MRW measurements. RNFLT measurements were obtained using a 12-degree circle scan centered on the optic disc. The diameter of the circle would be 3.5mm in eyes with standard corneal curvature. The circle scan is made up of 1536 A-scan points. BMO-MRW measurements were obtained using a 24-line radial scan of the optic nerve head centered on the BMO. The shortest distance from the BMO to the internal limiting membrane was measured at 48 data points around the optic nerve head. The software calculated an average measurement for global, temporal, supero-temporal, infero-temporal, nasal, supero-nasal, and infero-nasal RNFLT and BMO-MRW for each patient. The software also provided a quality score for each scan, ranging from 0 dB (poor) to 40 dB (excellent). Scans with quality score of 15 dB or less were excluded. All scans were manually reviewed by a reading center. If segmentation errors were identified, an attempt was made to correct them manually. If the segmentation error was considered uncorrectable, the scan was excluded from analysis.

Statistical Analysis

Variables were compared between glaucoma and normal groups using a two-tailed t test for continuous variables and a chi-squared test for categorical variables. The diagnostic accuracy of each OCT parameter was evaluated by constructing receiver operating characteristic (ROC) curves. ROC curves were plotted to demonstrate the tradeoff between the sensitivity and 1 – specificity. The area under the ROC curve (AUC) was used to assess the diagnostic accuracy of each parameter, with 1.0 representing perfect discrimination and 0.5 representing chance discrimination. For each parameter, we also calculated sensitivities at fixed specificities of 80% and 95%. We also investigated the impact of age on the diagnostic accuracy of RNFLT and BMO-MRW. This was formally investigated by including age as a covariate in a ROC regression model, as previously described.¹⁵ In addition, we also calculated ROC curve areas stratifying subjects by the median age of the sample.

As the dataset included data from both eyes from the same patient in some cases, a bootstrap resampling procedure was used to derive 95% confidence intervals and P-values, where the eye level clusters was considered as the unit of resampling. This procedure is commonly used to account for the presence of multiple correlated measurements within the same subject.^{15, 16} All analyses were performed using STATA (version 14.0, StataCorp, College Station, TX, USA).

Results

The study included 113 eyes from 81 patients suspected of having glaucoma. Of the 113 eyes included in the study, 87 (77%) were considered glaucoma suspect based solely on thin rim/large cup, 8 (7%) based solely on suspicious RNFL defects, and 18 (16%) based on both characteristics. Of the 113 eyes suspected of having glaucoma, 52 (46.0%) eyes had progressive glaucomatous optic nerve changes and were classified as having glaucoma and 61 (54.0%) eyes did not have progressive glaucomatous optic nerve changes and were classified as normal controls. Of the 52 eyes that had progressive glaucomatous optic nerve changes, 28 (54%) were considered progressing based solely on neuroretinal rim thinning/enlarging cup, 10 (19%) based solely on RNFL loss, and 14 (27%) based on both characteristics. Descriptive characteristics of the study population are presented in Table 1. No subjects developed an abnormal visual field during follow-up.

Table 1.

Characteristics of the study population of glaucoma suspects with and without glaucoma based on the presence of progressive glaucomatous optic nerve changes.

	Glaucomaa (52 eyes of 43 subjects)		Controlsb (61 eyes of 38 subjects)
	N	%	N	%	P valuec
Gender					0.07
Female	22	51.2	27	71.1
Male	21	48.8	11	29.9
	Mean	SD	Mean	SD
Age (years)	67.5	12.4	67.1	9.9	0.85
Time between earliest and most recent photos (years)	10.2	4.8	9.4	3.2	0.26
MD (dB)	−0.65	1.65	0.22	1.23	0.002
PSD (dB)	1.70	0.25	1.52	0.23	<0.001

^aDefined as glaucoma suspects with progressive glaucomatous optic nerve changes seen on stereoscopic optic disc photography.

^bDefined as glaucoma suspects without progressive glaucomatous optic nerve changes seen on stereoscopic optic disc photography.

^cP values were calculated with a chi-squared test for categorical variables or with a two-tailed t test for continuous variables.

MD = mean deviation; PSD = pattern standard deviation; dB = decibels

Table 2 contains the means for each of the RNFLT and BMO-MRW parameters. There was a statistically significant difference between the glaucomatous eyes and control eyes for each of the parameters evaluated. The mean ± SD global BMO-MRW was 212.4 ± 7.1 microns for glaucomatous eyes and 254.7±4.9 for control eyes (p<0.001). The mean ± SD global RNFLT was 74.8±1.4 microns for glaucomatous eyes and 91.4±1.0 for control eyes (p<0.001). The distributions of measurements for the mean global BMO-MRW and the mean global RNFLT in eyes with and without progressive glaucomatous optic nerve changes are shown in Figure 1.

Table 2.

Mean values, areas under receiver operating characteristic curve, and sensitivities at fixed specificities of 95% and 80% for each of the retinal nerve fiber layer thickness and Bruch’s membrane opening minimum rim width parameters.

Parameter	Mean±SD Glaucoma eyes (μm)	Mean±SD Control eyes (μm)	P Value	Area under ROC curve	Sensitivity at 95% Specificity	Sensitivity at 80% Specificity
BMO-MRW
Global	212.4±7.1	254.7±4.9	<0.001	0.75	0.40	0.65
Temporal	158.9±6.5	179.8±4.5	0.01	0.66	0.21	0.52
Supero-temporal	201.3±7.2	245.2±6.4	<0.001	0.75	0.29	0.67
Infero-temporal	207.4±8.3	276.6±5.9	<0.001	0.82	0.54	0.67
Nasal	233.7±6.8	280.7±7.1	<0.001	0.70	0.33	0.48
Supero-nasal	242±9.3	280.1±6.8	0.001	0.69	0.33	0.58
Infero-nasal	260.6±9.3	314.0±7.0	<0.001	0.75	0.25	0.67
RNFLT
Global	74.8±1.4	91.4±1.0	<0.001	0.89	0.60	0.77
Temporal	56.4±1.8	64.9±1.3	<0.001	0.69	0.37	0.54
Supero-temporal	102.4±2.8	126.2±2.3	<0.001	0.80	0.50	0.71
Infero-temporal	100.5±3.0	128.8±1.8	<0.001	0.87	0.54	0.71
Nasal	58.8±1.4	70.9±1.5	<0.001	0.77	0.31	0.62
Supero-nasal	78.3±3.0	100.5±2.2	<0.001	0.81	0.50	0.65
Infero-nasal	86.9±2.5	104.4±2.7	<0.001	0.73	0.27	0.50

BMO-MRW = Bruch’s membrane opening minimum rim width. RNFLT = retinal nerve fiber layer thickness. ROC = receiver operating characteristic.

Figure 1.

Histogram showing the distribution of measurements for global Bruch’s membrane opening minimum rim width (BMO-MRW) and global retinal nerve fiber layer thickness (RNFLT) in glaucoma and normal eyes. Diagnosis of glaucoma was based on previously documented history of stereophotographic optic disc progression occurring in eyes that were suspected of glaucoma at the time of optical coherence tomography imaging.

The areas under the ROC curves for discriminating glaucomatous from normal eyes were 0.89 (95% CI: 0.84–0.95) for global RNFLT and 0.75 (95% CI: 0.65–0.85) for global MRW (p=0.006). At 95% specificity, sensitivities were 60% for the global RNFLT parameter and 40% for the global BMO-MRW parameter. The ROC curves for global RNFLT and global BMO-MRW are shown in Figure 2. The areas under the ROC curves for the sectoral measurements (temporal, supero-temporal, infero-temporal, nasal, supero-nasal, infero-nasal) ranged from 0.66 to 0.82 for BMO-MRW and 0.69 to 0.87 for RNFLT (Table 2). Although, by definition, all eyes had normal visual fields at the time of imaging, some differences in MD and PSD were seen between healthy and glaucoma eyes (Table 1). However, ROC curve areas for SAP MD and PSD were only 0.65 (95% CI: 0.55 – 0.75) and 0.69 (0.59 – 0.79), respectively. Figure 2 also shows the ROC curve for SAP PSD.

Figure 2.

Receiver operating characteristic curves for global Bruch’s membrane opening minimum rim width (BMO-MRW), global retinal nerve fiber layer thickness (RNFLT), and standard automated perimetry pattern standard deviation (PSD) in discriminating glaucomatous and normal eyes. Diagnosis of glaucoma was based on previously documented history of stereophotographic optic disc progression occurring in eyes that were suspected of glaucoma at the time of optical coherence tomography imaging.

In a ROC regression model, age was significantly associated with the diagnostic accuracy of global BMO-MRW (p=0.003). ROC curve areas for BMO-MRW declined significantly for older age. Age was not significantly associated with diagnostic accuracy (ROC curve area) for global RNFLT (p=0.448). For patients older than the median age (68 years), the areas under the ROC curves for discriminating glaucomatous eyes from normal eyes were 0.89 (95% CI: 0.77–0.95) for global RNFLT and 0.63 (95% CI: 0.50–0.77) for global BMO-MRW (p<0.001). For patients with age equal to or younger than the median, the ROC curve areas were 0.88 (95% CI: 0.78–0.96) for global RNFLT and 0.86 (95% CI: 0.73–0.95) for global BMO-MRW (p=0.75).

Discussion

In our analysis RNFLT performed better than BMO-MRW at identifying pre-perimetric glaucomatous optic nerve damage in eyes suspected of having glaucoma. While previous studies have compared the performance of these two OCT parameters, their investigations were limited to comparisons involving eyes with clearly defined visual field loss and healthy controls who exhibited no signs suspicious for glaucoma.^7,9–11 In contrast, the present study investigated a population of individuals suspected of having glaucoma whose final diagnoses were determined by longitudinal observation. Such design provides information that is likely to be more relevant in clinical practice when OCT is being used as an ancillary diagnostic test to help identify disease in glaucoma suspects.

Both RNFLT and BMO-MRW measurements were significantly lower in glaucoma suspects who were found to have glaucomatous damage versus those who were ultimately classified as normal. However, the analysis of diagnostic accuracy revealed a superior performance of RNFLT parameters. While the area under the ROC curve for RNFLT was 0.89, the one for global BMO-MRW was only 0.75, which also translated into a significant difference in sensitivity at matched specificities. It is interesting to speculate why RNFLT would perform better than BMO-MRW in this circumstance, especially considering that previous studies involving patients with glaucomatous visual field loss have found similar accuracies for these parameters. In clinical practice, glaucoma suspects are usually those referred due to enlarged cups, as this is a relatively easy sign to detect during fundoscopic examination. In comparison, it is relatively uncommon to find patients who are referred as suspects due solely to the appearance of the RNFL, as fundoscopic examination of the RNFL is challenging and visualization is often difficult. Therefore, the clinical population of suspects tends to be enriched by subjects with eyes showing enlarged cups or suspicious rim thinning and many of those eyes prove to be ultimately normal on subsequent follow-up. In fact, most of the eyes in our study were initially classified as suspects based solely on suspicious appearance of the rim/cup (77%) versus only 7% who were classified as suspects due solely to the appearance of the RNFL. In 16%, the diagnosis of suspect was based on both characteristics. Therefore, in the presence of suspicious cupping, it is not surprising that an imaging test that evaluates the RNFL would provide more information to help define the diagnosis. In fact, a previous study comparing neuroretinal rim area assessment by confocal scanning laser ophthalmoscopy and RNFL assessment with scanning polarimetry reached similar conclusions as the present one.¹⁴

An interesting observation was that there was a marked influence of age on the diagnostic accuracy of BMO-MRW, with this parameter performing significantly worse for detecting glaucoma in older subjects compared to younger ones. Previous studies have shown greater proportional decline in BMO-MRW measurements compared to RNFL thickness as a result of aging.¹⁷ It is possible that these increased age-related losses may lead to smaller differences in BMO-MRW measurements between glaucomatous and normal populations at older age, explaining the decrease in discriminative ability. Another possible explanation is that the cup and lamina cribrosa are more mobile and influenced by IOP changes in younger patients than older patients.¹⁸ BMO-MRW measurements in younger eyes would therefore be more reflective of the increased IOP levels often seen in glaucoma, making it easier to differentiate them from normal eyes with normal IOP. This is a relevant finding, given that glaucoma is usually a disease of older subjects. It should be noted, however, that our sample size was relatively small to fully explore the effects of aging on diagnostic accuracy and these findings deserve additional investigation.

We used evidence of documented progressive disc change to separate glaucoma suspect patients in those who were disease positive versus disease negative. As all patients were required to have normal visual fields at the time of imaging, no other reference standard would be available to classify these patients in this situation. In the absence of visual field loss, it is very difficult if not impossible to be certain about the presence of glaucoma on a cross-sectional examination. Demonstrating a history of progressive glaucomatous changes to the optic nerve allowed us to reach a determination of which eyes had glaucoma based on previously documented deterioration versus those that were likely to be normal, given the absence of deterioration despite lack of treatment for many years. The use of progressive optic disc change as a reference standard, however, has some limitations. Demonstration of progressive optic disc change requires longitudinal follow-up and serial documentation of optic disc appearance, which may not be available for all patients. It might be argued that some suspects ultimately classified as normal could have had glaucomatous damage, but the follow-up time was insufficient to detect progression. Although it is unlikely that glaucoma patients would not progress or develop functional loss observed for almost 9 years without treatment, this possibility cannot be completely discarded. The requirement for no treatment was applied only to patients included in the control group to avoid any confounding effects of treatment in the assessment of progression in this group.

It is important to keep in mind that different reference standards may be required to assess the performance of imaging tests in different situations. Our design enabled the evaluation of their ability to diagnose disease in patients with suspicious disc appearance. In contrast, if the purpose of the study was to evaluate how these tests would perform for population-based or opportunistic screening, the study design would have to be different and our findings would not be directly applicable. For detecting patients with glaucomatous visual field loss in the general population for glaucoma screening, a study design such as the one used in previous studies enrolling a wide variety of patients with confirmed visual field loss would be more appropriate. An important message is that diagnostic studies should be designed to evaluate tests in the context where the test will likely be used in clinical practice. While our findings suggest that RNFLT is more useful than BMO-MRW when evaluating glaucoma suspects, there may be other situations where BMO-MRW is more useful than RNFLT. For example, one study showed that BMO-MRW had fewer false-positive results than RNFLT when evaluating tilted optic discs.¹⁹ In our study, only 12 (11%) of the 113 eyes had tilted discs. Unfortunately, this small number prevented us from being able to assess the impact of this condition on diagnostic accuracy. BMO-MRW may also be more useful when extensive peri-papillary atrophy or epiretinal membranes are present. Further work may identify other situations where these parameters may perform differently.

Our study has limitations. In order to reach a decision about the presence of glaucoma, we had to use photographic evidence of progressive optic disc damage. Therefore, identifying damage required a substantial amount of nerve tissue to have been lost in order to be identifiable by photographic comparison. It is possible that glaucoma patients with milder degrees of damage would therefore not be represented in the study, which would tend to overestimate the diagnostic accuracies of the parameters investigated. Further, using progressive optic disc damage as a reference standard may have biased the decision in favor of parameters assessing features that are more easily detected as changing over time. However, we expect that this would tend to favor optic disc topographic parameters rather than RNFL ones, as decisions about progression from optic disc photographs were more frequently based on progressive enlargement of the cup or thinning of the rim, rather than loss of RNFL over time. As an additional limitation, due to our inclusion and exclusion criteria, some populations were not represented in the study, such as subjects with high myopia. It is possible that the diagnostic performance of OCT parameters may behave differently in those subjects and further studies should be undertaken to clarify this.

In conclusion, our findings suggest that RNFLT is better able to identify pre-perimetric glaucomatous damage in glaucoma suspects than BMO-MRW. This may make RNFLT measures more useful than BMO-MRW measures for the practicing ophthalmologist when evaluating glaucoma suspects.

When evaluating glaucoma suspects in clinical practice, retinal nerve fiber layer thickness parameters may be better able to identify glaucomatous damage than Bruch’s membrane opening minimum rim width measurements.