Comparison of Cirrus OCT and Stratus OCT on the Ability to Detect Localized Retinal Nerve Fiber Layer Defects in Preperimetric Glaucoma

Abstract

Purpose.: To evaluate and compare the diagnostic ability of Cirrus and Stratus optical coherence tomography (OCT) to detect localized retinal nerve fiber layer (RNFL) defects in patients with normal standard automated perimetry.

Methods.: This study included 55 eyes of 55 subjects with preperimetric localized RNFL defects and 55 normal control eyes of 55 age- and sex-matched subjects. Areas under the receiver operating characteristic curves (AUROCs) were calculated and compared. Based on the internal normative database from each device, the sensitivity and specificity for detecting preperimetric localized RNFL defects were calculated.

Results.: There was no statistically significant difference between the AUROCs for the best parameters from the Cirrus OCT (inferior thickness, AUROC = 0.728) and Stratus OCT (7 o'clock sector, AUROC = 0.760; P = 0.477). The sensitivity of the Cirrus OCT parameters ranged from 21.0% to 87.1% and that of the Stratus OCT parameters ranged from 4.8% to 30.7%, with the criterion of abnormal at the 5% level. Based on the normative database, the highest Cirrus OCT sensitivity was obtained with the deviation-from-normal map (sensitivity 87.1% and specificity 61.8%), and the highest Stratus OCT sensitivity was obtained with the TSNIT thickness graph (sensitivity 30.7% and specificity 85.5%).

Conclusions.: There were no significant differences between the AUROCs for Cirrus and Stratus OCT, indicating that the two devices have similar diagnostic potentials in preperimetric glaucoma. After comparison with their normative databases, Cirrus OCT had generally higher sensitivities; however, this was largely at the cost of lower specificities than Stratus OCT.

Glaucoma is an optic neuropathy characterized by structural changes to the optic nerve head and retinal nerve fiber layer (RNFL), with corresponding functional changes, particularly visual field (VF) loss. RNFL loss is thought to precede measurable optic nerve head and visual field damage and is observed in 60% of eyes approximately 6 years before any detectable VF defects. ^1–4 Evaluation of RNFL damage is therefore of vital importance for the diagnosis of glaucoma in the early stages. 

Several techniques are currently available for detecting and quantifying RNFL damage, such as clinical examination, red-free fundus photography, and modern imaging devices. The optical properties of the RNFL have allowed recent advances in ocular imaging technology to obtain thickness measurements. ⁵ Optical coherence tomography (OCT) is a new technology that allows for quantitative assessment of RNFL thickness. It offers objective, real-time assessment of the RNFL within a very short time span at a single visit. ^6,7 The ability of Stratus OCT to provide quantitative and reproducible measurements of RNFL thickness parameters has been documented in previous studies. ^8–10 To date, several studies investigating the performance of Stratus OCT in glaucoma with manifest VF defect have shown promising results. ^11,12 However, it has been demonstrated that Stratus OCT has a relatively low sensitivity for identifying localized RNFL defects in preperimetric glaucoma. ¹³  

Spectral-domain OCT has recently been introduced, with improved image resolution, imaging speed, and sensitivity. ^14–16 Cirrus OCT, which has recently become commercially available, acquires OCT data approximately 70 times faster and with better resolution (5 μm vs. 8–10 μm axial resolution in tissue), compared to Stratus OCT technology. Thus, based on these technological improvements, it can be expected that Cirrus OCT has superior performances in the early stages of glaucoma, particularly in preperimetric glaucoma, in which Stratus OCT has generally low sensitivity. 

To the best of our knowledge, there have been no reports addressing the diagnostic performance of Cirrus OCT in preperimetric glaucoma. Therefore, this study was designed to evaluate and compare the diagnostic abilities of Cirrus and Stratus OCT for detecting localized RNFL defects in preperimetric glaucoma. 

Materials and Methods

Eyes with preperimetric localized RNFL defects and normal control eyes meeting the eligibility criteria were consecutively enrolled from the Glaucoma Clinic of Seoul National University Hospital during the period from May 2008 to October 2008. In cases in which both eyes of a subject were eligible for the study, only one eye was randomly chosen for inclusion. When one eye was preperimetric and the other was normal, the preperimetric eye was selected for this study. The study adhered to the tenets of the Declaration of Helsinki and was approved by the Institutional Review Board of the Seoul National University Hospital. Informed consent was obtained from all subjects. 

All subjects underwent a complete ophthalmic examination including visual acuity, manifest refraction, intraocular pressure (IOP) measurements by Goldmann applanation tonometry, slit lamp examination, gonioscopy, dilated fundus examination with a 78-D lens, color disc photography, red-free RNFL photography (VX-10; Kowa Optimed, Tokyo, Japan), Swedish interactive thresholding algorithm (SITA) 30-2 perimetry (Humphrey field analyzer II; Carl Zeiss Meditec, Dublin, CA), Cirrus spectral-domain OCT (Carl Zeiss Meditec), and Stratus OCT (Carl Zeiss Meditec). 

All subjects had a best-corrected visual acuity of 20/40 or better, spherical equivalent refractive error within ±5.00 D, astigmatism within ±3.00 D, an open anterior chamber angle, good quality of red-free photography, and reliable VFs. The inclusion criteria were normal standard automated perimetry (SAP) results in at least two tests. Normal VF indices were defined as a mean deviation and pattern standard deviation within 95% confidence limits and a glaucoma hemifield test result within normal limits. 

We excluded eyes with a history of uveitis, ocular surgery other than cataract extraction, or diseases that may affect the peripapillary area where OCT measurements are obtained (e.g., large peripapillary atrophy, chorioretinal coloboma, and peripapillary staphyloma). Eyes with consistently unreliable VFs (defined as false negative >33%; false positive >33%; and fixation losses >20%) were excluded from the study. 

Subjects were assigned to either the preperimetric localized RNFL defect group or the normal control group. The preperimetric localized RNFL defect group was defined as those having a localized wedge-shaped RNFL defect clearly visible by red-free fundus photography with normal SAP results. The normal control group was defined as those having an IOP ≤ 21 mm Hg with no history of increased IOP, absence of glaucomatous disc appearance, no visible RNFL defect according to red-free RNFL photography, and a normal SAP result. Absence of glaucomatous disc appearance was defined as an intact neuroretinal rim without peripapillary hemorrhages, notches, or localized pallor. 

Red-Free RNFL Photography

Red-free RNFL photography was acquired with a digital fundus camera after maximum pupil dilation. Sixty degree, wide-angle views of the optic disc, carefully focused on the retina using the built-in split-line focusing device, were obtained and reviewed on an LCD monitor. ^12,17–20 Localized RNFL defects by red-free RNFL photography were determined when their width at a 1-disc-diameter distance from the edge of the disc was larger than that of a major retinal vessel, diverging in an arcuate or wedge shape and reaching the edge of the disc (Fig. 1A). ¹⁷  

Figure 1.

 

Defining localized RNFL defects. (A) A localized RNFL defect shown by red-free RNFL photography was determined when the width at a 1-disc-diameter distance from the edge of the disc was larger than a major retinal vessel, diverging in an arcuate or wedge shape and reaching the edge of the disc. (B) On the Cirrus OCT deviation-from-normal map, we defined wedge-shaped defects (arrows) radiating from the optic nerve head as Cirrus OCT RNFL defects.

Two experienced observers classified the red-free RNFL photographs as showing no visible RNFL defect, a localized defect, diffuse atrophy, or ambiguous information. Only subjects classified by both observers as having either no visible RNFL defect or localized defect were included in the study. Also, the observers measured the topographic parameters of the localized RNFL defects. This method has been described elsewhere in detail. ^12,13,18 In brief, red-free RNFL images were evaluated independently by the two observers in a random order and masked fashion, without knowledge of the clinical information and OCT results. Clock-face circles were drawn around the optic nerve head on the red-free photograph, the diameter and location of which corresponded as closely as possible to the circle displayed in the video mode of the RNFL thickness analysis report. The two points where the borders of the defect met the circle were then identified, and the directional angles of these border points with respect to the center of the circle were then measured. The angle between the two points was defined as the angular width of the RNFL defect. The directional angle was assessed in a clockwise direction in right eyes and in a counterclockwise direction in left eyes, with the temporal equator set at 0°. Defects were assigned a clock hour location according to the measured angles. Based on the localized RNFL defect locations described as a clock hour on the red-free RNFL photograph, we performed a topographic comparison with the Cirrus OCT RNFL measurements. 

Cirrus Spectral-Domain OCT

Subjects were measured with the Optic Disc Cube 200 × 200 program of the Cirrus HD-OCT model 4000 (software version 3.0). All Cirrus OCT scans were obtained after achieving pupillary dilation, followed by Stratus OCT Fast RNFL scan. The Optic Disc Cube 200 × 200 program obtains 200A-scans from 200 linear B-scans evenly distributed in a 6-mm² area centered over the optic nerve. 

Image quality of Cirrus OCT scans were assessed by two experienced examiners masked to the clinical information. The minimum acceptable signal strength score was 6. Moreover, the examiners assessed subjectively the quality of the image and evaluated the en face image for eye movements. Any subject with less than satisfactory Cirrus OCT image quality was excluded. 

Cirrus OCT extracts from the data cube 256 A-scan samples along the path of the calculation circle. Based on the RNFL layer boundaries in the extracted circle scan image, the Cirrus OCT calculates the RNFL thickness at each point along the calculation circle. Using these data, Cirrus OCT provides the 12-clock-hour thicknesses, four quadrant thicknesses, a global 360° average thickness, and TSNIT thickness profiles. For each parameter, the Cirrus OCT software provides a classification (within normal limits, borderline, or outside normal limits) based on comparison with an internal normative database. A parameter is classified as outside normal limits if its value falls lower than the 99% confidence interval (CI) of the healthy, age-matched population. A borderline result indicates that the value is between the 95% and 99% CI, and a within-normal-limits result indicates that the value is within the 95% CI. Segments of the TSNIT thickness graph located below the yellow band (outside of the 95% normal limit) and in the red band (outside of the 99% normal limit) were defined as OCT RNFL defects by TSNIT graph at the 5% and 1% level, respectively. 

The Cirrus OCT printout also provides deviation-from-normal maps that are derived from superpixel average thickness measurements and report a statistical comparison against the normal thickness range for each superpixel, overlaid on the OCT fundus image. These maps apply the yellow and red colors of the age-matched normative data to superpixels whose average thickness falls in the yellow and red normal distribution percentiles (i.e., 1%–5% and <1% of normal distribution percentiles, respectively). 

On the deviation-from-normal map, we defined wedge-shaped defects shown in yellow or red and radiating from the optic nerve head as Cirrus OCT RNFL defects (Fig. 1B). The minimum size of qualified defect in the deviation-from-normal map was when the defect width at a 1-disc-diameter distance from the edge of the disc was larger than that of a major retinal vessel. The interpretation was performed by two observers independently in a masked fashion. The presence of Cirrus OCT RNFL defects was determined by a consensus agreement between the two observers. 

On the deviation-from-normal map, clock-hour locations of Cirrus OCT RNFL defects were determined using a clock-face circle of a diameter equal to that of the calculation circle shown in the map. Clock hours were assessed in a clockwise direction for right eyes and in a counterclockwise direction for left eyes, with the temporal sector set at 9 o'clock. 

Stratus OCT

The fast RNFL algorithm was used to obtain RNFL thickness measurements with the Stratus OCT. Subjects were measured after achieving pupillary dilation to a minimum diameter of 5 mm. An internal fixation target was used, since it offers better reproducibility. ⁸ Data were analyzed with the Stratus system software (version 4.0). Three scan images were acquired from each subject, with each image consisting of 256 A-scans along a 3.4-mm-diameter circular ring around the optic disc. These values were averaged to yield 12 clock-hour thicknesses, four quadrant thicknesses, and a global average RNFL thickness measurement. For each parameter, the Stratus OCT software provides a classification (within normal limits, borderline, or outside normal limits) based on comparison with an internal normative database. In Stratus OCT, TSNIT thickness profiles were also obtained, which displayed thicknesses at each A-scan location along the path of the scan circle. In Stratus OCT, OCT RNFL defects by TSNIT graph were defined in the same way as in the Cirrus OCT. 

Quality assessments of Stratus OCT scans were performed by two experienced examiners masked to the subject's identity and the results of the other tests. Satisfactory quality was defined as: (1) well-focused images, (2) the presence of a centered circular ring around the optic disc, and (3) a signal strength ≥ 6 (10 = maximum). Any subject with less than satisfactory Stratus OCT image quality was excluded. 

Visual Field Testing

The SITA standard of the Humphrey Field Analyzer II 750 was performed, using the central full-threshold program 30-2. Glaucomatous VF loss was defined as the consistent presence of a cluster of three or more nonedge points on the pattern deviation plot with a probability of occurring in <5% of the normal population, with one of these points having the probability of occurring in <1% of the normal population; a pattern standard deviation with P < 5%; or a glaucoma hemifield test result outside normal limits. VF defects had to be repeatable on at least two consecutive tests. 

Data Analysis

All analyses were performed with two commercial programs (SPSS for Windows Version 15.0; SPSS Inc., Chicago, IL, or MedCalc ver. 10.0; MedCalc Software, Mariakerke, Belgium). P < 0.05 was considered statistically significant. 

Statistical Analysis Based on the Quantitative OCT RNFL Thickness Data.

Student's t-tests were used to evaluate RNFL thickness differences between preperimetric glaucoma and normal control eyes. We used receiver operating characteristic (ROC) curves to describe the diagnostic ability of each OCT parameter to differentiate preperimetric glaucoma from normal eyes. An area under the ROC curve (AUROC) of 1.0 represents perfect discrimination, whereas an AUROC of 0.5 represents chance discrimination. The method of DeLong et al. ²¹ was used to compare AUROCs. Sensitivity was assessed fixed at 80% and 95% of specificity. 

Statistical Analysis Based on the Internal Normative Database.

Diagnostic categorization (within normal limits, borderline, or outside normal limits) provided by each device after comparison with its respective normative database was evaluated, and the sensitivity, specificity, and likelihood ratio (LRs) were calculated. The methods regarding the LRs have been described elsewhere in detail. ^10,22,23 According to the classification suggested by Jaeschke et al., ²³ LRs higher than 10 or lower than 0.1 would be associated with large effects on posttest probability, LRs from 5 to 10 or 0.1 to 0.2 would be associated with moderate effects, LRs from 2 to 5 or 0.2 to 0.5 would be associated with small effects, and LRs closer to 1 would be insignificant. The 95% CIs for LRs were calculated by the method of Simel et al. ²⁴  

Results

During the enrollment period, this study initially involved 170 eyes of 170 subjects (62 eyes with preperimetric RNFL defects and 108 normal control eyes). Of these 170 eyes, 5 (2.9%) with unacceptable Cirrus OCT scans and 14 (8.2%) with unacceptable Stratus OCT scans were excluded from further analysis, leaving a sample of 151 eyes of 151 subjects (55 eyes with preperimetric RNFL defects and 96 normal control eyes). Of the 96 normal control eyes, 55 eyes of 55 subjects who were age- and sex-matched with subjects of preperimetric RNFL defects were selected for the analysis. Matching for age was performed by randomly selecting one control subject within 2 years of age for each preperimetric RNFL defect subject. Therefore, the final study sample included 110 eyes of 110 subjects (55 eyes of 55 patients with preperimetric RNFL defects and 55 healthy control eyes of 55 age- and sex-matched subjects). 

Subject Basic Characteristics

Demographic characteristics are detailed in Table 1. Since the normal control subjects were selected with age- and sex-matching, there was no significant difference in the mean age between the normal control subjects (53.4 ± 10.6 years; range, 25–74) and those with preperimetric RNFL defects (54.1 ± 10.4; range, 26–74). The sex of the subjects, refraction, and central corneal thickness were similar between the two groups. No significant differences were found between the two groups with respect to SAP mean deviation or pattern standard deviation. 

Table 1.

 

Demographic Characteristics of Study Subjects

	Preperimetric Glaucoma	Normal Control	P
Eyes, n	55	55
Age, y	54.1 ± 10.4	53.4 ± 10.6	0.730*
Gender ratio (male/female)	30:25	30:25	1.000†
Laterality (right/left)	27:28	32:23	0.444†
Spherical equivalent, D	−1.83 ± 3.26	−1.35 ± 2.71	0.576*
Central corneal thickness, μm	535.1 ± 27.3	541.5 ± 38.7	0.435*
Humphrey C30-2 threshold visual field
Mean deviation, dB	−0.74 ± 0.96	−0.36 ± 1.69	0.260*
Pattern standard deviation, dB	1.85 ± 0.39	1.69 ± 0.30	0.100*

*  Student's t-test.

†  χ² test.

RNFL Thickness Measurements and ROC Analysis

The Cirrus and Stratus OCT parameters computed in the control and preperimetric glaucoma groups are presented in Table 2. Statistically significant differences were found in the Cirrus OCT parameters of the 6-, 7-, 10-, 11-, and 12-o'clock sectors, inferior and superior quadrants, and average RNFL thickness. Likewise, for the Stratus OCT parameters, statistically significant differences were found in the 5-, 6-, 7-, 10-, and 11-o'clock sectors, inferior and superior quadrants, and average RNFL thickness. 

Table 2.

 

Cirrus and Stratus OCT Parameters

	Cirrus OCT			Stratus OCT
	Preperimetric Glaucoma	Normal Control	P *	Preperimetric Glaucoma	Normal Control	P *
Average	86.1 ± 10.2	94.0 ± 9.0	<0.001	99.5 ± 12.5	108.8 ± 10.7	<0.001
Quadrant
Superior	106.7 ± 19.2	119.1 ± 12.7	<0.001	121.2 ± 19.0	133.3 ± 15.8	<0.001
Temporal	65.8 ± 12.2	70.1 ± 11.2	0.057	77.5 ± 15.0	81.9 ± 12.2	0.091
Inferior	105.9 ± 17.5	119.2 ± 15.5	<0.001	122.2 ± 20.7	139.2 ± 17.3	<0.001
Nasal	66.3 ± 8.8	67.6 ± 10.7	0.479	77.3 ± 16.0	80.9 ± 16.0	0.243
Clock hours
12 Superior	109.9 ± 26.7	121.1 ± 23.6	0.021	124.1 ± 26.6	132.5 ± 23.2	0.080
11	107.4 ± 25.9	127.5 ± 19.7	<0.001	123.5 ± 21.3	142.7 ± 18.8	<0.001
10	74.1 ± 15.7	81.2 ± 14.8	0.016	88.0 ± 18.0	96.2 ± 16.4	0.015
9 Temporal	54.8 ± 10.8	55.8 ± 10.1	0.636	64.7 ± 12.9	64.8 ± 10.6	0.974
8	68.5 ± 17.1	73.2 ± 16.1	0.142	79.9 ± 19.9	84.9 ± 15.9	0.151
7	115.1 ± 27.3	137.1 ± 21.7	<0.001	127.1 ± 29.5	151.8 ± 19.8	<0.001
6 Inferior	115.6 ± 24.9	127.9 ± 25.2	0.012	133.7 ± 26.7	149.6 ± 25.4	0.002
5	87.1 ± 19.2	92.7 ± 17.4	0.115	105.8 ± 22.5	116.0 ± 21.2	0.016
4	61.5 ± 9.3	62.4 ± 11.5	0.642	76.5 ± 17.3	79.4 ± 17.1	0.380
3 Nasal	59.4 ± 9.0	57.1 ± 10.1	0.201	66.4 ± 15.4	67.0 ± 16.8	0.845
2	77.7 ± 15.3	83.2 ± 16.8	0.077	88.9 ± 21.5	96.1 ± 19.2	0.067
1	103.0 ± 24.8	109.0 ± 23.9	0.200	115.9 ± 25.7	124.7 ± 23.1	0.061

*  Comparison was performed using Student's t-test.

Table 3 indicates ROC curve areas and sensitivities at fixed specificities for all OCT parameters. The two Cirrus OCT parameters with the largest AUROCs were inferior thickness (0.728) and 7 o'clock sector (0.726). Similarly, the two Stratus OCT parameters with the largest AUROCs were the 7-o'clock sector (0.760) and the 11-o'clock sector (0.749). When all corresponding parameters were compared, there were no significant differences between the AUROCs for Cirrus and Stratus OCT. Moreover, no significant difference was found between AUROCs for the best parameters from the Cirrus OCT and Stratus OCT (P = 0.477; Fig. 2). 

Table 3.

 

AUROCs and Sensitivities at Fixed Specificities for the Cirrus and Stratus OCT Parameters

	AUROC (SE)			Sensitivity/Specificity
				Specificity ≥ 80.0%		Specificity ≥ 95.0%
	Cirrus OCT	Stratus OCT	P *	Cirrus OCT	Stratus OCT	Cirrus OCT	Stratus OCT
Average	0.716 (0.049)	0.697 (0.050)	0.469	43.6/80.0	43.6/80.0	14.5/96.4	27.3/96.4
Quadrant
Superior	0.699 (0.050)	0.692 (0.050)	0.800	29.1/81.8	49.1/80.0	5.5/96.4	9.1/96.4
Temporal	0.631 (0.053)	0.632 (0.053)	0.977	38.2/80.0	29.1/80.0	3.6/96.4	1.8/96.4
Inferior	0.728 (0.048)	0.724 (0.048)	0.872	36.4/81.8	32.7/80.0	12.7/96.4	18.2/96.4
Nasal	0.516 (0.055)	0.555 (0.055)	0.375	23.6/80.0	30.9/81.8	9.1/96.4	9.1/98.2
Clock hours
12 Superior	0.625 (0.053)	0.605 (0.054)	0.424	41.8/80.0	30.9/81.8	5.5/96.4	1.8/96.4
11	0.716 (0.049)	0.749 (0.047)	0.284	36.4/80.0	49.1/80.0	16.4/96.4	10.9/96.4
10	0.620 (0.053)	0.637 (0.053)	0.552	21.8/83.6	36.4/80.0	9.1/96.4	7.3/96.4
9 Temporal	0.529 (0.055)	0.529 (0.055)	0.989	23.6/81.8	23.6/81.8	3.6/96.4	1:8/98.2
8	0.599 (0.054)	0.613 (0.054)	0.635	32.7/80.0	32.7/80.0	3.6/96.4	5.5/96.4
7	0.726 (0.048)	0.760 (0.046)	0.196	47.3/80.0	58.2/80.0	18.2/96.4	5.5/96.4
6 Inferior	0.636 (0.053)	0.644 (0.053)	0.755	30.9/80.0	34.5/80.0	14.5/96.4	12.7/96.4
5	0.578 (0.055)	0.607 (0.054)	0.283	25.5/80.0	27.3/81.8	12.7/96.4	10.9/96.4
4	0.485 (0.055)	0.552 (0.055)	0.159	23.6/83.6	27.3/83.6	3.6/96.4	1.8/96.4
3 Nasal	0.577 (0.055)	0.506 (0.055)	0.158	29.1/81.8	27.3/81.8	1.8/96.4	5.5/96.4
2	0.587 (0.054)	0.597 (0.054)	0.795	29.1/80.0	27.3/80.0	10.9/96.4	9.1/98.2
1	0.557 (0.055)	0.603 (0.054)	0.120	23.6/80.0	18.2/80.0	3.6/98.2	9.1/96.4

*  Comparison was performed using the method of DeLong et al. ²¹

Figure 2.

 

ROC curves of the best parameters from the Cirrus and Stratus OCT.

Sensitivity, Specificity, and Likelihood Ratios Based on the Internal Normative Database

In the preperimetric glaucoma group, 48 eyes showed one localized RNFL defect, and 7 eyes showed two localized defects, according to red-free fundus photography. There were 62 localized RNFL defects in total. Localized RNFL defects were observed in the superotemporal (33 RNFL defects) and inferotemporal (29 RNFL defects) regions. 

Table 4 presents the sensitivity and specificity for the overall OCT parameters. When the normative database was used, the sensitivity of Cirrus OCT was generally higher than that of Stratus OCT. Based on the normative database, the sensitivity of the various Cirrus OCT parameters ranged from 21.0% to 87.1%, and that of the Stratus OCT parameters ranged from 4.8% to 30.7%, with the criterion of abnormal at the 5% level. The two Cirrus OCT parameters with the highest sensitivity were the deviation-from-normal map at the 5% level (sensitivity 87.1% and specificity 61.8%) and the TSNIT thickness graph at the 5% level (sensitivity 79.0% and specificity 65.5%). The two Stratus OCT parameters with the highest sensitivity were the TSNIT thickness graph at the 5% level (sensitivity 30.7% and specificity 85.5%) and ≥1 clock hour abnormal at the 5% level (sensitivity 22.6% and specificity 85.5%). Based on the normative database, the specificity of the Stratus OCT parameters (range, 85.5%–100.0%) was generally higher than that of the Cirrus OCT parameters (range, 61.8%–100.0%). Figure 3 shows the Cirrus and Stratus OCT results from three cases. 

Table 4.

 

Sensitivity and Specificity for Cirrus and Stratus OCT Parameters Based on the Internal Normative Database

	Cirrus OCT		Stratus OCT
	Sensitivity (%)	Specificity (%)	Sensitivity (%)	Specificity (%)
≥1 Clock hour
Abnormal at the 5% level	56.5 (43.3–69.0)	80.0 (67.0–89.6)	22.6 (12.9–35.0)	85.5 (73.3–93.5)
Abnormal at the 1% level	16.1 (8.0–27.7)	96.4 (87.5–99.5)	3.2 (0.5–11.2)	100.0 (93.5–100.0)
≥1 Quadrant
Abnormal at the 5% level	33.9 (22.3–47.0)	98.2 (90.2–99.7)	11.3 (4.7–21.9)	98.2 (90.2–99.7)
Abnormal at the 1% level	19.4 (10.4–31.4)	100.0 (93.5–100.0)	4.8 (1.1–13.5)	100.0 (93.5–100.0)
Average RNFL thickness
Abnormal at the 5% level	21.0 (11.7–33.2)	100.0 (93.5–100.0)	4.8 (1.1–13.5)	100.0 (93.5–100.0)
Abnormal at the 1% level	6.5 (1.8–15.7)	100.0 (93.5–100.0)	1.6 (0.3–8.7)	100.0 (93.5–100.0)
TSNIT thickness graph
Abnormal at the 5% level	79.0 (66.8–88.3)	65.5 (51.4–77.8)	30.7 (19.6–43.7)	85.5 (73.3–93.5)
Abnormal at the 1% level	41.9 (29.5–55.2)	94.6 (84.9–98.8)	9.7 (3.7–19.9)	100.0 (93.5–100.0)
Deviation-from-normal map
Abnormal at the 5% level	87.1 (76.1–94.2)	61.8 (47.7–74.6)	—	—
Abnormal at the 1% level	67.7 (54.7–79.1)	92.7 (82.4–97.9)	—	—

Figure 3.

 

Cases of preperimetric glaucoma. (A) The localized RNFL defect shown by red-free photography (arrowheads) was detected on the deviation-from-normal map (arrow) and circular diagrams of Cirrus OCT. This defect was also identified on the TSNIT graph and circular diagrams of Stratus OCT. The defect detected by Cirrus and Stratus OCT showed excellent topographic agreement with the defect location seen by red-free photography. (B) The localized RNFL defect by red-free photography (arrowheads) was detected only in the deviation-from-normal map of Cirrus OCT (arrow). This case suggests that the deviation-from-normal map is the most sensitive parameter for identifying early RNFL damage. (C) Red-free photography and OCT results in a case of normal control eye. Red-free photography showed no visible RNFL defect. However, multiple defects (arrows) were observed on the Cirrus OCT deviation-from-normal map. It is presently unknown whether these defects are true subclinical RNFL losses or false-positive findings. The true nature of these defects may be defined only through a longitudinal assessment.

Table 5 presents LRs with their 95% CIs for the Cirrus and Stratus OCT parameters. For the Cirrus OCT overall parameters, the LRs of outside normal limits ranged from 4.44 to infinity, and those of borderline results ranged from 0.63 to 7.98. Likewise, for the Stratus OCT overall parameters, the LRs of outside normal limits were infinity and those of borderline results ranged from 1.33 to 3.55. The LRs of within normal limits results ranged from 0.21 to 0.67 for the Cirrus OCT overall parameters and from 0.81 to 0.91 for the Stratus OCT overall parameters. 

Table 5.

 

Likelihood Ratios for Cirrus and Stratus OCT Parameters

	Cirrus OCT			Stratus OCT
	Within Normal Limits	Borderline	Outside Normal Limits	Within Normal Limits	Borderline	Outside Normal Limits
Average	0.78 (0.68–0.90)	Infinity (NA)	Infinity (NA)	0.95 (0.89–1.01)	Infinity (NA)	Infinity (NA)
Quadrant
Superior	0.73 (0.62–0.86)	Infinity (NA)	Infinity (NA)	0.96 (0.92–1.01)	Infinity (NA)	Infinity (NA)
Temporal	0.98 (0.95–1.02)	Infinity (NA)	Infinity (NA)	1.00 (1.00–1.00)	Infinity (NA)	Infinity (NA)
Inferior	0.76 (0.65–0.89)	9.00 (1.18–68.66)	Infinity (NA)	0.87 (0.79–0.97)	Infinity (NA)	Infinity (NA)
Nasal	0.98 (0.95–1.02)	Infinity (NA)	Infinity (NA)	0.96 (0.90–1.04)	2.00 (0.19–21.42)	Infinity (NA)
Clock hours
12 Superior	0.85 (0.77–0.95)	Infinity (NA)	Infinity (NA)	0.96 (0.90–1.04)	3.00 (0.32–27.96)	Infinity (NA)
11	0.66 (0.54–0.81)	9.00 (2.19–36.95)	Infinity (NA)	0.96 (0.90–1.04)	3.00 (0.32–27.96)	Infinity (NA)
10	0.93 (0.86–1.00)	Infinity (NA)	Infinity (NA)	0.96 (0.92–1.01)	Infinity (NA)	Infinity (NA)
9 Temporal	0.98 (0.92–1.04)	2.00 (0.19–21.42)	Infinity (NA)	1.00 (1.00–1.00)	Infinity (NA)	Infinity (NA)
8	0.96 (0.92–1.01)	Infinity (NA)	Infinity (NA)	1.00 (1.00–1.00)	Infinity (NA)	Infinity (NA)
7	0.83 (0.73–0.95)	7.00 (0.89–55.02)	Infinity (NA)	0.87 (0.78–0.98)	7.00 (0.89–55.02)	Infinity (NA)
6 Inferior	0.94 (0.83–1.07)	1.00 (0.21–4.74)	4.00 (0.46–34.66)	0.96 (0.92–1.01)	Infinity (NA)	Infinity (NA)
5	0.87 (0.78–0.98)	5.00 (0.60–41.42)	Infinity (NA)	0.93 (0.86–1.00)	Infinity (NA)	Infinity (NA)
4	1.04 (0.99–1.09)	0.00	Infinity (NA)	1.04 (0.97–1.12)	0.33 (0.04–3.11)	Infinity (NA)
3 Nasal	1.02 (0.98–1.06)	Infinity (NA)	0.00	1.00 (0.95–1.05)	1.00 (0.06–15.59)	Infinity (NA)
2	0.95 (0.89–1.01)	Infinity (NA)	Infinity (NA)	0.85 (0.75–0.96)	8.00 (1.04–61.83)	Infinity (NA)
1	0.87 (0.76–0.99)	4.00 (0.89–17.99)	Infinity (NA)	0.96 (0.90–1.04)	2.00 (0.19–21.42)	Infinity (NA)
Overall parameters
≥1 clock hour	0.54 (0.40–0.74)	2.46 (1.26–4.81)	4.44 (1.02–19.37)	0.91 (0.76–1.08)	1.33 (0.59–3.01)	Infinity (NA)
≥ 1 quadrant	0.67 (0.56–0.81)	7.98 (1.04–61.03)	Infinity (NA)	0.90 (0.82–0.99)	3.55 (0.41–30.80)	Infinity (NA)
TSNIT graph	0.32 (0.19–0.54)	1.28 (0.75–2.16)	7.69 (2.46–24.01)	0.81 (0.67–0.99)	1.44 (0.65–3.22)	Infinity (NA)
Deviation-from-normal map	0.21 (0.11–0.41)	0.63 (0.33–1.19)	9.31 (3.57–24.31)	—	—	—

 NA, not applicable.

Topographic Relationship between Red-free RNFL Photography and Cirrus OCT Deviation-from-Normal Map

A subset of 22 eyes with a single localized RNFL defect at 7 and/or 8 o'clock in the red-free RNFL photography showed 19 defects at the same location by the Cirrus OCT deviation-from-normal map. Of these 22 subset eyes, 6 had OCT RNFL defects at the superior hemifield, where red-free RNFL photography revealed normal RNFL configuration. Similarly, a subset of 26 eyes having a single localized RNFL defect at 10 and/or 11 o'clock in the red-free RNFL photography showed 24 defects at the same location by Cirrus OCT. Of these 26 subset eyes, 10 RNFL defects that were not seen in the RNFL photography were visualized at the inferior hemifield in the deviation-from-normal maps. 

Angular Widths of RNFL Defects and Sensitivity of Both Devices

The sensitivity of the Cirrus and Stratus OCT parameters was closely related to the angular widths of the preperimetric RNFL defects (Table 6). For the preperimetric RNFL defects with angular widths ≤10°, the sensitivities of Cirrus and Stratus OCT were 55.6% and 5.6%, respectively, for the TSNIT graph at the 5% level. For preperimetric RNFL defects with angular widths ≥ 20°, higher sensitivity was noted on the TSNIT graph parameters, resulting in the Cirrus OCT sensitivity of 83.3% and the Stratus OCT sensitivity of 50.0%. Table 6 summarizes the sensitivity of each device's parameters according to the angular widths of the preperimetric RNFL defects. 

Table 6.

 

Angular Widths of Preperimetric RNFL Defects and Sensitivities for Cirrus and Stratus OCT Parameters

Angular Width (Degrees)	Number of RNFL Defects	TSNIT Thickness Graph*		Clock Hour Map†		Deviation-from- Normal Map‡
		Sensitivity of Cirrus OCT (%)	Sensitivity of Stratus OCT (%)	Sensitivity of Cirrus OCT (%)	Sensitivity of Stratus OCT (%)	Sensitivity of Cirrus OCT (%)
<5.0	9	44.4 (14.0–78.6)	0.0 (0.0–33.8)	0.0 (0.0–33.8)	0.0 (0.0–33.8)	44.4 (14.0–78.6)
5.0–9.9	9	66.7 (30.1–92.1)	11.1 (1.8–48.3)	44.4 (14.0–78.6)	0.0 (0.0–33.8)	88.9 (51.7–98.2)
10.0–14.9	20	90.0 (68.3–98.5)	35.0 (15.5–59.2)	55.0 (31.6–76.9)	20.0 (5.9–43.7)	95.0 (75.1–99.2)
15.0–19.9	12	91.7 (61.5–98.6)	41.7 (15.3–72.3)	75.0 (42.8–94.2)	41.7 (15.3–72.3)	100.0 (73.4–100.0)
>20.0	12	83.3 (51.6–97.4)	50.0 (21.2–78.8)	91.7 (61.5–98.6)	41.7 (15.3–72.3)	91.7 (61.5–98.6)
Total	62	79.0 (66.8–88.3)	30.7 (19.6–43.7)	56.5 (43.3–69.0)	22.6 (12.9–35.0)	87.1 (76.1–94.2)

*  TSNIT thickness graph abnormal at the 5% level.

†  ≥1 Clock hour abnormal at the 5% level.

‡  Deviation-from-normal map abnormal at the 5% level.

Discussion

The present study was designed with the main objective of evaluating the diagnostic abilities of Cirrus OCT and Stratus OCT for detecting preperimetric localized RNFL defects. We confirmed that the Cirrus and Stratus OCT have similar diagnostic potentials for discriminating between healthy eyes and eyes with a localized RNFL defect (according to red-free fundus photography) but no VF defect (according to SAP). 

In our study, several measures of diagnostic accuracy were performed, including ROC analysis. ROC curve areas for earlier versions of OCT have ranged from 0.79 to 0.94, depending on the parameters and characteristics of the included subjects. ^10,25–30 In previous studies, RNFL thickness in the inferior region often had the best ability to discriminate normal eyes from eyes with early to moderate glaucoma, with sensitivities between 64% and 79% for specificities of 90% or higher. ^10,25,26,30 In our study, the best Cirrus and Stratus OCT parameters with the largest AUROCs were inferior thickness (AUROC = 0.728) and 7-o'clock sector (AUROC = 0.760), respectively. However, direct comparisons with previous studies may be limited because the characteristics of the study population are quite different. 

This study has demonstrated that, based on the respective normative database, the sensitivity of the Cirrus OCT is generally higher than that of the Stratus OCT. However, when matched for the fixed specificity, the sensitivities of the two devices were almost identical. Therefore, the Cirrus OCT parameters only have higher sensitivities when the normative database is used and at the cost of lower specificities. Our results suggest that the higher image resolution by the Cirrus OCT may not translate directly to a superior diagnostic performance for early glaucoma detection. Although the two devices have similar diagnostic performances, the Cirrus OCT provides new information regarding the peripapillary RNFL status through the deviation-from-normal map. However, this map requires a human subjective assessment, which may limit its applicability in clinical practice. Therefore, further software development may be needed to use this obvious improvement in detection that derives from looking at the RNFL over a broad area. 

This study has shown that both devices are still inadequate to detect preperimetric RNFL defects, especially ones that are smaller than 10°. Considering the sensitivity/specificity analysis is heavily dependent on the severity of glaucomatous damage, the relatively low sensitivity may be related to the population evaluated, which only included early glaucoma patients. It is expected that the Cirrus and Stratus OCT may have higher sensitivities for the eyes with manifest VF defects. 

In the subset analysis, a total of 16 RNFL defects that were not seen in the RNFL photography were detected in the Cirrus OCT deviation-from-normal maps. It is not clear whether these defects are false-positive findings or whether they represent true RNFL losses. Because the true nature of these OCT defects may be confirmed by a follow-up assessment, the performance of OCT in these cases should be re-evaluated through a prospective longitudinal study. Thus, if a definite glaucomatous change occurs in the case in which OCT had previously detected an abnormality, OCT may predict the future development of glaucoma in these questionable cases. 

The LRs provide a direct estimate of how much a test result will change the odds of having a disease. In our study, LRs for outside-normal-limits results from both devices were generally associated with moderate to large effects on the posttest probability of disease. However, LRs for within-normal-limits results were generally associated with small changes in probability, suggesting that these parameters would be of limited value in excluding the presence of disease. Test results for some parameters had LRs of infinity, which indicates that a particular test result was not found in any of the healthy subjects. 

A limitation of this study is the exclusion of the ambiguous group. The present study used a case-control design, including patients with an easily identifiable localized RNFL defect (cases) and normal subjects with a perfectly normal RNFL (controls). However, in clinical practice, glaucoma may appear without a prominent RNFL defect. Since this patient does not fall into either one of these two categories, diagnostic assessment in clinical practice can be much different from what the design of this study can accomplish. ³¹ Therefore, further studies are needed to determine the diagnostic accuracy of OCT in those types of damage. 

In conclusion, there were no significant differences between the AUROCs for Cirrus and Stratus OCT, indicating that the two devices have similar diagnostic potentials in preperimetric glaucoma. Among the various Cirrus and Stratus OCT parameters, the deviation-from-normal map and the TSNIT thickness graph had the highest sensitivity. After comparison with their normative databases, Cirrus OCT had generally higher sensitivities; however, this was largely at the cost of lower specificities than those of the Stratus OCT. Further studies may be needed to develop new parameters that will increase the diagnostic performance for the detection of early glaucomatous damage and its progression over time. 

Footnotes

 Disclosure: J.W. Jeoung, None; K.H. Park, None