Comparison of 10-2 and 24-2 Perimetry to Diagnose Glaucoma Using OCT as an Independent Reference Standard

Abstract

Purpose:

To compare the performance of the 10–2 versus the 24–2 standard automated perimetry (SAP) for the diagnosis of glaucoma, using optical coherence tomography (OCT) as an independent standard for glaucomatous damage.

Design:

Cross-sectional study.

Participants:

1,375 pairs of 10–2 and 24–2 SAP tests from 569 eyes of 339 subjects were used for the analysis. 440 (77%) eyes had a diagnosis of glaucoma, and 129 (23%) eyes were normal. All participants underwent 10–2 and 24–2 SAP within 30 days.

Methods:

Glaucomatous severity was quantified based on OCT macula ganglion cell layer (mGCL) and circumpapillary retinal nerve fiber layer (cpRNFL). The area under the receiver operating characteristic curve (AUC) was used to compare 10–2 and 24–2 metrics for discriminating healthy eyes from those of glaucoma, at different levels of disease severity.

Main outcome measures:

Areas under the ROC curves (AUC) and sensitivities at fixed specificities of 80% and 95%.

Results:

The overall AUC for mean deviation (MD) for the 24–2 test (0.808) was significantly higher than that of the 10–2 (0.742; P < 0.001). When compared at different stages of disease, the 24–2 test performed generally better than the 10–2, notably in earlier stages of disease. For early damage (1^st quartile), the 24–2 MD had an AUC of 0.658 versus 0.590 for 10–2 MD (P=0.018). For advanced damage (4^th quartile), corresponding values were 0.954 vs. 0.903 (P=0.013). Similar trends were observed when glaucoma severity was defined based on structural macular damage with mGCL thickness.

Conclusion:

The 24–2 SAP test had better diagnostic accuracy compared to the 10–2 for detecting equivalent levels of glaucomatous damage, as measured by quantitative assessment of RNFL and macula by OCT.

PRÉCIS

The accuracy of the standard perimetry 24–2 test was superior to that of the 10–2 test to detect glaucomatous damage quantified by optical coherence tomography assessment of retinal nerve fiber layer and macula.

Glaucoma has classically been described as initially impacting the peripheral visual field, with the more central visual field affected only in advanced stages. However, a growing body of evidence has suggested that glaucomatous damage may affect the macula even in early stages, which could be missed without targeted investigation of the centralmost area of the field of vision.^1–4

Standard automated perimetry (SAP) is the primary tool for diagnosing and monitoring visual field loss associated with glaucoma.^5,6 The 24–2 strategy of the Humphrey Field Analyzer (HFA, Carl-Zeiss Meditec, Dublin, CA) assesses 54 points within the central 24 degrees of the visual field, providing patterns of damage that can be readily recognizable by clinicians as glaucomatous defects. However, only a few of those points represent the macula – an area where more than 30% of the retinal ganglion cells reside⁷ – thus potentially not providing enough spatial resolution for defects present in this region. As such, recent studies have reported that a 68-point evaluation of the 10-degree central region, the 10–2 test, may detect central defects that are missed by the 24–2 SAP.^3,8–10 Given the potential for macular damage, some authors have proposed that the 10–2 test should be performed instead of, or at least in addition to, the 24–2 even in early stages of glaucoma.^4,9,11 Other studies suggest that the use of repeated testing with the 10–2 SAP might only burden the healthcare system and the patient, indicating that there is no clinical benefit to using the 10–2 in patients with early glaucoma.^12–15

So far, investigations comparing the diagnostic performance of the 24–2 and 10–2 tests in patients with glaucoma have been limited to selected groups (i.e., glaucoma suspects or early glaucoma), or without taking into account the possible effect of disease severity. If one method should replace the other, more information needs to be obtained on how the tests perform in the whole spectrum of the disease. Also, previous studies have suffered from a lack of an independent reference standard to define glaucomatous damage, a critical issue when comparing these two functional tests.^8,9 Optical coherence tomography (OCT) provides objective and reproducible quantitative assessment of neural tissue loss from both the circumpapillary retinal nerve fiber layer (cpRNFL) and the macular area and may serve as a reference standard to quantify glaucomatous damage. Therefore, this study aimed at evaluating the diagnostic performance of the 24–2 versus the 10–2 tests using OCT as an independent quantifier of glaucomatous damage, allowing comparison of the performance of these tests to diagnose specific levels of disease severity.

METHODS

This was a cross-sectional study with participants enrolled in a prospective, longitudinal study designed to evaluate structure and function in glaucoma. The study adhered to the Health Insurance Portability and Accountability Act, and all methods complied with the tenets of the Declaration of Helsinki guidelines for human subject research. The Institutional Review Board from Duke University approved the study, and all participants provided written informed consent.

During follow-up, participants underwent a comprehensive ophthalmologic examination, including review of medical history, visual acuity, slit lamp biomicroscopy, intraocular pressure measurement, gonioscopy, stereoscopic optic disc photography, dilated fundoscopic examination, SAP using 24–2 and 10–2 Swedish interactive threshold algorithm (SITA) standard (Carl Zeiss Meditec, Inc., Dublin, CA), and Spectralis spectral-domain OCT (Software version 5.8, Heidelberg Engineering, GmbH, Dossenheim, Germany) images and data. All subjects were required to have best corrected visual acuity of 20/40 or better, spherical refraction within ±5.0 D, cylinder correction <3.0 D, and open angles on gonioscopy. Subjects were excluded if they had any ocular or systemic disease that could affect the visual field or optic nerve in at least one of the eyes, other than glaucoma.

For inclusion in the study, participants were initially selected based on the presence of glaucomatous optic neuropathy (GON) using simultaneous stereoscopic optic disc photographs. Features indicating structural damage to the optic disc and RNFL, such as excavation, localized RNFL defects, and/or neuroretinal rim thinning, were used as the sole indicators of glaucomatous damage. Eyes with GON had the amount of glaucomatous damage then quantified based on OCT, as described below. Normal control subjects had normal optic discs and were also required to have intraocular pressures (IOP) of 22 mmHg or less with no history of increased IOP in both eyes. These visually healthy participants were recruited from patients’ spouses or relatives and were also recommended by ophthalmologists and optometrists from the Comprehensive Ophthalmology clinics at Duke Eye Center treating the population. Of note, standard definitions of the disease based on functional losses (e.g., presence of repeatable glaucomatous visual field defects, such as arcuate scotomas and nasal steps) were not used for a diagnosis of glaucoma since they would potentially bias the comparison between the two tests under investigation.

Standard Automated Perimetry

All eyes included were required to have reliable 10–2 and 24–2 SITA Standard perimetry tests performed within 30 days of each other. The SAP SITA 24–2 strategy tests 54 test locations within the central 24° of the visual field (30° nasally), evenly distributed with 6° separation; the SAP SITA 10–2 strategy evaluates 68 test point locations evenly distributed with 2° separation in the central 10-degrees. For both strategies, a test was considered unreliable if there were more than 33% fixation losses or more than 15% false positives. The global indices mean deviation (MD), pattern standard deviation (PSD), as well as the percentage of abnormal points (P<5%) in the total deviation (TD) plot were used for evaluation of the diagnostic accuracies of the visual function tests. All tests were reviewed for the presence of artifacts including eyelid or rim artifacts, or fatigue effects.

Spectral-Domain OCT

OCT images were acquired with the Spectralis spectral-domain OCT to measure the thickness of the cpRNFL and the macular ganglion cell layer (mGCL).¹⁶ The device uses a dual-beam spectral-domain OCT and a confocal laser scanning ophthalmoscope that emits a superluminescent diode light with a center wavelength of 870 nm and an infrared scan to simultaneously provide images of ocular microstructures.

cpRNFL measurements were obtained using the 3.45-mm diameter peripapillary circle scan, centered at the optic disc. For each circle scan, the global cpRNFL thickness was automatically calculated by the built-in software as the average thickness of all 768 points around the optic nerve head.¹⁷ Thickness measurements of the mGCL were obtained from an 8×8 posterior pole grid (each square or ‘superpixel’ measuring 3° in width), as described in detail previously.^16,18,19 The global thickness for each macular layer was calculated as the average thickness of all 64 sectors in the posterior pole region and was used in the analysis. All images were reviewed to ensure good quality, with signal strength greater than 15 dB. Macular scans with one or more of the cells in the 8×8 grid with missing data were also excluded from the analysis.

Statistical analysis

Receiver operating characteristic (ROC) curves were used to assess and compare the ability of the 10–2 and 24–2 SAP metrics (i.e., MD, PSD and percentage of TD points) for diagnosing glaucomatous optic neuropathy. The ROC curve is a plot of the sensitivity (true positive rate) and 1 - specificity (false positive rate), which describes the range of possible operating characteristics for a test.²⁰ They are frequently used to evaluate the accuracy of diagnostic tests when results are non-binary.^21,22 The area under the ROC curve (AUC) was then calculated and used to summarize the diagnostic accuracy of each parameter evaluated. The maximum AUC of 1.0 implies perfect differentiation between diseased and non-diseased, while an AUC of 0.5 represents chance discrimination.

Since we were particularly interested in the diagnostic performance of the 24–2 and 10–2 tests in different stages of the disease, glaucoma eyes were divided into subgroups of disease severity for analysis. Because there are no agreed-upon criteria for the use of OCT in staging levels of glaucomatous loss, the glaucoma cohort was stratified into quartiles of structural loss, as given by the cpRNFL or mGCL thicknesses. The lower (1^st) quartile (thicker measures) indicated early structural loss due to glaucoma, while the 4^th quartile (thinner measures) implied advanced disease. The subgroups were then compared to the same healthy cohort of normal controls (i.e., the healthy cohort was not stratified). ROC curves were adjusted for age at the time of testing in all models. Sensitivities at fixed specificities of 95% and 80% were reported for each parameter at each level of severity. To account for the fact that both eyes of the same participant were used for analyses, a bootstrap resampling procedure with 500 replications was used to calculate P-values and confidence intervals (CI) using the cluster of data for the subject as the unit of resampling. As measurements from both eyes of the same subject are likely to correlate, this procedure has been used in previous studies to adjust standard errors for the presence of multiple correlated measurements from the same unit.²⁰

Demographics and clinical characteristic were compared between diagnostic groups using appropriate statistical methods. Generalized estimating equations (GEE) were used to account for the fact that more than one test for the same eye was included in the sample. Categorical variables were compared using the Chi-square test. All statistical analyses were performed using Stata version 17.0 (StataCorp LLC, College Station, TX, USA).

RESULTS

The study included 1,375 pairs of 10–2 and 24–2 tests from 569 eyes of 339 participants. Four hundred forty (77%) eyes of 266 participants had a diagnosis of glaucoma, while 129 (23%) eyes of 73 participants were normal control eyes. Mean ages of normal subjects and glaucoma subjects were 64.9 ± 9.7 years versus 72.0 ± 10.6 years, respectively (P<0.001). 178 (52.5%) of all subjects were female. Table 1 shows demographic and clinical characteristics of the participants included in the study. Supplemental Figure 1 (available at https://www.ophthalmologyglaucoma.com) shows the distribution of 10–2 MD, 24–2 MD, cpRNFL, and mGCL values in the normal and glaucoma groups.

Table 1.

Baseline and clinical characteristics of participants and eyes in the normal and glaucoma groups.

	Normal 129 eyes of 73 subjects	Glaucoma 440 eyes of 266 subjects	P value
Age at baseline, years	64.9±9.7	72.0±10.6	<0.001a
Gender, n (%)
Female	50 (68.5%)	128 (48.1%)	0.002b
Race, n (%)
African American	25 (34.3%)	71 (26.7%)
Non-African American	48 (65.8%)	195 (73.3%)	0.204b
cpRNFL, μm	94.1±10.9	72.7±15.4	<0.001c
mGCL, μm	31.0±2.6	26.1±3.8	<0.001c
MD 10–2, dB*	−0.17 (−1.08; 0.45)	−2.23 (−6.81; −0.57)	<0.001c
MD 24–2, dB *	−0.15 (−1.15; 0.57)	−3.10 (−7.46; −1.08)	<0.001c
PSD 10–2, dB *	1.23 (1.08; 1.42)	2.00 (1.38; 8.68)	0.083c
PSD 24–2, dB *	1.60 (1.38; 1.99)	3.70 (2.12; 8.94)	0.003c
Abnormal TD points 10–2, %	2.0 (0.0; 7.0)	20.0 (4.0; 38.0)	<0.001c
Abnormal TD points 24–2, %	2.0 (0.0; 7.0)	19.0 (7.0; 34.0)	<0.001c

cpRNFL= circumpapillary RNFL; mGCL= GCL macula; MD= mean deviation; PSD= pattern standard deviation; TD = total deviation.

Data are presented as mean ± standard deviation, unless otherwise noted.

^*Median (Interquartile Range)

^aT-test

^bChi-square

^cGeneralized Estimating Equations

Overall, the 24–2 test demonstrated higher diagnostic performance as compared to the 10–2, across multiple indices when adjusted for age. The overall AUC for MD for the SAP 24–2 test was 0.808 versus 0.742 for the 10–2 (P<0.001). The AUC values for PSD and percentage of abnormal TD points were also higher for 24–2 (Table 2).

TABLE 2.

Areas under the receiving operator characteristic curves (AUC) for discriminating glaucoma from normal subjects for the mean deviation (MD), pattern standard deviation (PSD), and percentage of abnormal total deviation (TD) points, adjusted for age at time of testing.

SAP Parameter	10–2 AUC	24–2 AUC	P value
MD	0.742 (0.670–0.799)	0.808 (0.755–0.852)	<0.001
PSD	0.732 (0.661–0.817)	0.784 (0.716–0.846)	0.026
Abnormal TD Points	0.717 (0.638–0.772)	0.798 (0.744–0.843)	<0.001

SAP= standard automated perimetry; AUC= area under ROC curve; MD= mean deviation; PSD= pattern standard deviation; TD= total deviation

^*Numbers in parenthesis represent 95% confidence interval

Table 3 shows AUCs for 10–2 and 24–2 parameters to detect different quartiles of glaucomatous damage as measured by OCT cpRNFL and mGCL thickness. Quartiles were labeled so that the 1^st or lowest quartile (i.e., thicker cpRNFL or mGCL) denoted earlier damage, while the 4^th or highest quartile (i.e., thinner values) represented more advanced disease. As expected, higher AUCs were obtained for increasing levels of damage. However, important differences between the two tests were noted, with AUCs generally higher for the 24–2 compared to the 10–2 test. Differences were greater for earlier stages of damage and were present even when damage was measured by OCT mGCL thickness. For example, for detection of eyes with early damage (1^st quartile) based on cpRNFL thickness, AUCs for the 24–2 MD and 10–2 MD were 0.658 and 0.590, respectively (P=0.018). When severity was defined based on mGCL thickness, corresponding AUCs were 0.720 and 0.608 (P<0.001), respectively. For detection of advanced damage (4^th quartile), AUCs for 24–2 MD and 10–2 MD were 0.954 and 0.903, respectively (P=0.013). Similarly, for detection of advanced damage as defined by mGCL thickness, corresponding AUCs were 0.961 and 0.938, respectively; P=0.162. Figures 2 and 3 show ROC curves for different levels of disease damage as measured by cpRNFL and mGCL, respectively. Very similar patterns were observed when diagnostic accuracies were investigated using PSD or percentage of abnormal points on the TD plot (Table 3 and supplemental Figures 4 and 5, available at https://www.ophthalmologyglaucoma.com).

TABLE 3.

Areas under the receiver operator characteristic curves (AUCs) for the mean deviation (MD), pattern standard deviation (PSD), and percentage of abnormal total deviation (TD) points of the 10–2 and 24–2 visual field tests, stratified by quartiles of structural damage measured by the circumpapillary retinal nerve fiber layer (cpRNFL) and macular ganglion cell layer (mGCL) compared to the same healthy cohort. Higher quartiles indicate greater structural damage.

	MD			PSD			Abnormal TD points
	10–2	24–2	P value	10–2	24–2	P value	10–2	24–2	P value
cpRNFL Quartile
1st (>82 μm)	0.590	0.658	0.018	0.579	0.638	0.097	0.557	0.650	0.001
2nd (72.5–82 μm)	0.717	0.792	0.004	0.672	0.740	0.027	0.673	0.775	0.001
3rd (62–72.5 μm)	0.752	0.823	0.003	0.778	0.824	0.074	0.737	0.813	0.002
4th (<62 μm)	0.903	0.954	0.013	0.896	0.933	0.100	0.897	0.950	0.022
mGCL Quartile
1st (>28 μm)	0.608	0.720	<0.001	0.582	0.689	0.003	0.563	0.702	<0.001
2nd (26–28 μm)	0.679	0.723	0.070	0.663	0.688	0.469	0.651	0.709	0.037
3rd (24–26 μm)	0.744	0.830	0.004	0.760	0.821	0.024	0.722	0.832	0.001
4th (<24 μm)	0.938	0.961	0.162	0.922	0.937	0.394	0.931	0.948	0.374

MD= mean deviation; PSD= pattern standard deviation; TD= total deviation; cpRNFL= circumpapillary retinal nerve fiber layer; mGCL= macular ganglion cell layer

FIGURE 2.

Receiver operating characteristic (ROC) curves for standard automated perimetry mean deviation (MD) values of the 10–2 and 24–2 visual field tests for quartiles of circumpapillary retinal nerve fiber layer (cpRNFL) thickness compared to the same healthy cohort. Higher quartiles indicate greater structural loss.

FIGURE 3.

Receiver operating characteristic (ROC) curves for standard automated perimetry mean deviation (MD) values of the 10–2 and 24–2 visual field tests for quartiles of macula ganglion cell layer (mGCL) thickness compared to the same healthy cohort. Higher quartiles indicate greater structural loss.

Table 4 shows sensitivities at fixed specificities of 80% and 95% according to different subgroups of disease severity. At 80% specificity, the sensitivity was 92% for 24–2 MD versus 82% for 10–2 MD (P=0.016) for detection of eyes with advanced RNFL loss (4^th quartile). As expected, a much lower sensitivity was seen for both tests when detecting eyes with early structural loss (1^st quartile of RNFL thickness), with sensitivities at 80% specificity of 40% for 24–2 MD and 33% for 10–2 MD (P=0.122). Similar patterns were observed when damage was assessed by mGCL thickness and also for different visual field parameters.

TABLE 4.

Sensitivities at fixed specificities of 80% and 95% for the 10–2 and 24–2 tests at quartiles of circumpapillary retinal nerve fiber layer (cpRNFL) and macular ganglion cell layer (mGCL) loss compared to the same healthy cohort. Higher (4^th) quartiles indicate greater structural damage.

	MD			PSD
	10–2	24–2	P value	10–2	24–2	P value
1st Quartile cpRNFL
95% specificity	19% (10%-30%)	21% (11%-29%)	0.669	12% (4%-34%)	18% (8%-30%)	0.355
80% specificity	33% (22%-46%)	40% (27%-51%)	0.122	20% (7%-49%)	27% (12–46%)	0.453
4th Quartile cpRNFL
95% specificity	71% (61%-80%)	85% (75%-92%)	<0.001	72% (55%-89%)	79% (63%-93%)	0.118
80% specificity	82% (69%-91%)	92% (83%-98%)	0.016	77% (63%-93%)	88% (76%-96%)	0.083
1st Quartile mGCL
95% specificity	24% (16%-35%)	30% (20%-45%)	0.174	16% (7%-34%)	26% (13%-43%)	0.043
80% specificity	33% (23%-45%)	50% (40%-60%)	<0.001	22% (11%-50%)	38% (23%-56%)	0.027
4th Quartile mGCL
95% specificity	79% (69%-88%)	87% (77%-94%)	0.083	78% (62%-91%)	80% (62%-91%)	0.641
80% specificity	88% (77%-94%)	93% (87%-97%)	0.157	83% (66%-96%)	89% (77%-96%)	0.194

MD= mean deviation; PSD= pattern standard deviation; cpRNFL= circumpapillary retinal nerve fiber layer; mGCL= macular ganglion cell layer

^*Numbers in parenthesis represent 95% confidence interval

Figure 6 illustrates a case of glaucomatous damage, with cpRNFL thickness of 68μm and mGCL thickness of 26μm. A superior nasal step can be seen on 24–2 SAP with abnormal PSD, which corresponds to inferior thinning as detected on OCT cpRNFL. However, the 10–2 test looks normal. Figure 7 depicts a case of a glaucomatous eye with advanced cpRNFL structural damage, with thickness of 50μm (most severe quartile), and milder macular damage, with mGCL thickness of 27μm. Superior and inferior arcuate defects are seen on 24–2 SAP as well as on the 10–2 test.

FIGURE 6.

Glaucomatous eye with a mean global circumpapillary retinal nerve fiber layer (cpRNFL) thickness of 68μm (3^rd quartile of loss) and mean macular ganglion cell layer (mGCL) thickness of 26μm (3^rd quartile of loss). A superior nasal step is detected by the 24–2 test with abnormal pattern standard deviation, but the 10–2 test is normal.

FIGURE 7.

Glaucomatous eye with mean global circumpapillary retinal nerve fiber layer (cpRNFL) thickness of 50μm (4^th quartile of loss) and mean macular ganglion cell layer (mGCL) thickness of 27μm (2^nd quartile of loss). Both the 24–2 and 10–2 tests show abnormal results with superior and inferior arcuate defects.

DISCUSSION

In the present study, we compared diagnostic accuracies of the 24–2 and 10–2 SAP for detecting glaucomatous damage based on structural loss as determined by OCT. The 24–2 test exhibited higher accuracy throughout the spectrum of severity, as determined by both OCT cpRNFL and macular thickness measurements. These findings suggest that the 24–2 should generally be the preferred method of SAP testing for diagnostic purposes.

We evaluated the global indices MD and PSD, as well as the percentage of abnormal TD points from both 10–2 and 24–2 SAP tests, while incorporating OCT as an unbiased reference of disease severity. As expected, diagnostic performances of all indices were higher with increasing disease severity. For greater disease severity, represented by the 4^th quartile of cpRNFL thickness, AUCs for both 24–2 and 10–2 were large, of 0.954 and 0.903, respectively (P=0.013). This is to be expected. With such significant loss, both tests would detect damage easily, although the 24–2 outperformed the 10–2. A similar trend was found in subjects within the 4^th quartile of mGCL thickness, with AUCs of 0.961 vs. 0.938, respectively, although the difference between the two tests here was not statistically significant (P=0.162). However, at less profound levels of neural loss, a greater difference between the diagnostic performances of 10–2 and 24–2 tests was observed, favoring the latter. At the 1^st quartile of cpRNFL thickness, indicating earliest damage, the AUC was 0.658 for the 24–2 but only 0.590 for the 10–2 MD (P=0.018). Interestingly, this was also noted even when structural damage was measured in the macula, with corresponding AUCs of 0.720 and 0.608, respectively (P<0.001), for detecting eyes at the 1^st quartile of mGCL thickness. This may seem intriguing, as one would expect that mild macular damage would be better picked up by the 10–2 test. These findings suggest that when milder macular damage is observed on OCT, significant damage has most likely already occurred in the periphery. In support of this, one can see that the diagnostic accuracies of visual field parameters were generally higher for the same subgroup of disease severity in the macula compared to cpRNFL, at the highest and lowest quartiles of structural measurements (Table 3).

To further investigate detection of structural macular damage with the 24–2 and 10–2 tests, we also analyzed the diagnostic performance of the TD points obtained solely within the central 10 degrees from fixation in the 24–2 SAP. Although all the central points in the 24–2 pattern do not have an exact match in the 10–2, this corresponds approximately to the 12 innermost points, which are within the central 10-degree area sampled by the 68 points in the 10–2 test. Interestingly, similar results were observed (Supplemental Table 5, available at https://www.ophthalmologyglaucoma.com). AUCs for the central points in the 24–2 test still outperformed the 10–2 test. This is a surprising result for which we do not have a clear explanation, but it may perhaps be related to different normative databases used for the estimation of sensitivity thresholds as part of the Bayesian SITA strategy in 24–2 compared to that used in 10–2. In a previous study, Wu et al derived PSD values from the central 12 locations of the 24–2 and compared it to the PSD values of the 10–2 visual field, finding a similar performance of both methods to detect central visual field abnormalities.²³ These results were subsequently confirmed by other investigators.^13,14,24

Comparison of the 10–2 and 24–2 visual fields tests has been a topic of recent debate, given the important association of central vision with quality of life.^25,26 Recent studies have suggested that the 10–2 detects as much as a third of macular damage that would be missed by the 24–2.^3,8–10 However, although these studies have reported higher frequencies of central glaucomatous damage than previous investigations, it is possible that their convenience samples may have included a higher percentage of patients that were tested by the 10–2 only because of concerns of macular damage raised by the clinician, and thus may represent a biased sample not fully representative of the overall glaucoma population. To avoid this potential bias, all eyes in our study were tested with both the 10–2 and 24–2 as part of a pre-established longitudinal protocol which recruited consecutive subjects from a glaucoma clinic.

There is data showing that visual fields are not routinely carried out for management of glaucoma with the frequency recommended by most professional organizations.^27,28 Therefore, a routine implementation of 10–2 visual testing in addition to the conventional 24–2 test could put additional strain on an already imperfect system, without providing a clear benefit. Alternatives have been proposed, such as the 24–2C test, which incorporates 10 additional test points within the central 10 degrees, increasing the sampling of the central visual field. However, studies are needed to assess the diagnostic performance of this new method at different stages of the disease. Future studies should also determine the value of adding points to the visual field test in comparison to a standard combination of OCT imaging and conventional 24–2 testing. As OCT imaging of multiple topographic regions can be performed very quickly, one can easily image macula and RNFL in the same session to optimize detection of damage.

It is important to emphasize that, in comparison to the most advanced stages of RNFL loss, AUCs for both perimetric tests were much more modest in eyes with mild to moderate glaucomatous damage, suggesting great difficulty for detecting eyes with early structural damage. This is not a surprising finding and has been shown previously by several other studies. The lack of sensitivity of SAP for detecting early structural damage is likely related to the properties of perimetric testing, such as the use of a logarithmic scale for assessing and reporting sensitivity thresholds, as pointed out by several studies,^29–31 but may also be related to other factors such as redundancy and test-retest variability. Given the low sensitivity of both 24–2 and 10–2 SAP for detecting early OCT damage, the findings of our study highlight the importance of imaging assessment for early glaucoma diagnosis.

The findings of this study should not be interpreted as if the 10–2 SAP adds no value to the clinical management of glaucoma. In fact, the 10–2 test is the test of choice for monitoring eyes with advanced damage where no or little visual sensitivity is observed outside the central 5° or 10° of the central visual field. The 10–2 test may also be useful in monitoring defects that threaten fixation.^5,6 Further studies should be conducted to clarify when it would be clinically useful to switch to or add 10–2 testing in the management of glaucoma.⁹

Our study had limitations. Our results were derived from a single institute cohort of patients with primary open-angle glaucoma, so they may not be generalizable to a population with other forms of glaucoma or different demographics. Additionally, we only used summary metrics to evaluate diagnostic accuracy.^24,32,33 It could be argued that clinicians would evaluate the whole printout instead of only summary parameters when assessing these tests for diagnostic purposes. However, a recent study by Wu et al³⁴ showed that subjective assessment of visual field printouts by glaucoma experts performed no better than global indices for detecting glaucoma damage. In fact, subjective expert assessment of 10–2 SAP was worse than global metrics, likely from the fact that 10–2 visual fields generally do not exhibit the clear patterns of visual field loss that clinicians are accustomed to. As another limitation, our assessment of structural damage also relied on global OCT metrics. However, such metrics have been widely demonstrated to be highly reproducible and to have strong diagnostic accuracy.^35,36 Importantly, the same metrics were used to provide a reference standard for an unbiased comparison of 24–2 and 10–2 testing. Finally, it should be noted that although we recruited consecutive patients with a diagnosis of GON based solely on structural evaluation of the optic disc in an attempt to avoid favoring either visual field test, it may be argued that it is not possible to fully eliminate biases. It is possible that patients may have been originally referred for care because of particular test results on the 24–2 or 10–2 tests. However, our study replicates what clinicians are often faced on their first evaluation of subjects for glaucoma in a clinical setting. The only way to fully eliminate such biases would be to apply the two perimetric tests under a population-based study, which is likely unfeasible.

In conclusion, this study found that SAP testing with the 24–2 pattern was generally superior to 10–2 testing for detecting glaucoma damage, as measured by OCT. Our findings suggest that the 24–2 test should in general be the preferred method for visual function assessment in glaucoma.

Supplementary Material

1

SUPPLEMENTAL FIGURE 1. Frequencies of standard automated perimetry mean deviation (MD) values of the 10–2 and 24–2 visual field tests and OCT circumpapillary retinal nerve fiber layer (cpRNFL) and macular ganglion cell layer (mGCL) thicknesses, stratified by diagnosis of glaucoma versus normal.

2

SUPPLEMENTAL FIGURE 4. Receiver operating characteristic (ROC) curves for standard automated perimetry pattern standard deviation (PSD) values of the 10–2 and 24–2 visual field tests for quartiles of retinal nerve fiber layer (cpRNFL) thickness compared to the same healthy cohort. Higher quartiles indicate greater structural loss.

3

SUPPLEMENTAL FIGURE 5. Receiver operating characteristic (ROC) curves for the percentage of abnormal standard automated perimetry total deviation (TD) points values of the 10–2 and 24–2 visual field tests for quartiles of circumpapillary retinal nerve fiber layer (cpRNFL) thickness compared to the same healthy cohort. Higher quartiles indicate greater structural loss.