Glaucoma Diagnostic Accuracy of OCT with a Normative Database to Detect Diffuse Retinal Nerve Fiber Layer Atrophy: Diffuse Atrophy Imaging Study Jin Wook Jeoung, 1,2 Seok Hwan Kim, 1,3 Ki Ho Park, 1,2 Tae-Woo Kim, 1,4 and Dong Myung Kim 1,2 From the 1 Department of Ophthalmology, Seoul National University College of Medicine, Seoul, Korea; 2 Department of Ophthalmology, Seoul National University Hospital, Seoul, Korea; 3 Department of Ophthalmology, Seoul National University Boramae Hospital, Seoul, Korea; and 4 Department of Ophthalmology, Seoul National University Bundang Hospital, Seongnam, Korea. Submitted for publication October 10, 2010; revised April 3 and May 21, 2011; accepted June 8, 2011. Disclosure: J.W. Jeoung, None; S.H. Kim, None; K.H. Park, None; T.-W. Kim, None; D.M. Kim, None Corresponding author: Seok Hwan Kim, Department of Ophthalmology, Seoul National University Boramae Hospital, #425 Shindaebang-dong, Dongjak-gu, Seoul 156-707, Korea; xcski@hanmail.net. PURPOSE. To investigate the diagnostic accuracy of Stratus optical coherence tomography (OCT) with its internal normative database to detect diffuse retinal nerve fiber layer (RNFL) atrophy in glaucoma subjects. METHODS. One hundred two eyes of 102 glaucoma patients with diffuse RNFL atrophy and 102 healthy eyes of 102 agematched subjects were enrolled in the Diffuse Atrophy Imaging Study. Two experienced observers graded RNFL photographs of diffuse RNFL atrophy eyes using a four-level grading system. The diagnostic performances for detecting diffuse RNFL atrophy were examined according to visual field results and RNFL photograph grading. RESULTS. Using a criterion of abnormal at the 5% level, the overall sensitivity of the Stratus OCT parameters ranged from 61.5% to 84.5%, and the overall specificity ranged from 90.2% to 99.0%. For mild, moderate, and severe diffuse RNFL atrophy, the superior quadrant had a sensitivity of 41.0%, 83.3%, and 100.0%, respectively, and the inferior quadrant had a sensitivity of 35.0%, 88.5%, and 100.0%, respectively. The highest likelihood ratios were obtained at the 11 and 12 o clock sectors for superior RNFL and the 6 and 7 o clock sectors for inferior RNFL. CONCLUSIONS. OCT with a normative database can detect diffuse RNFL atrophy with moderate sensitivity and high specificity. Because the sensitivity of Stratus OCT is closely related to the degree of diffuse RNFL atrophy and the visual field results, OCT with an internal normative database should be evaluated with prudence, especially in the early stage of glaucoma with diffuse RNFL atrophy. (Invest Ophthalmol Vis Sci. 2011;52:6074 6080) DOI: 10.1167/iovs.10-6703 Since Hoyt et al. 1 initially described retinal nerve fiber layer (RNFL) atrophy in glaucoma, the evaluation of RNFL has been of vital importance for diagnosing glaucomatous optic neuropathy. Several techniques have been suggested to detect and quantify RNFL damage, such as red-free RNFL photography and modern imaging devices. Red-free RNFL photography provides a permanent and accurate record of the RNFL and can serve as a qualitative or semiquantitative reference of RNFL damage. 2 Recently, quantitative methods of imaging RNFL have been increasingly used to diagnose glaucomatous optic neuropathy. Of these, optical coherence tomography (OCT) can provide high-resolution, crosssectional images that allow in vivo measurement of tissue thickness. 3,4 The ability of Stratus OCT to provide quantitative measurements of RNFL thickness has been documented in previous studies. 5 7 Clinically, RNFL atrophy can be divided into localized atrophy (wedge-shaped defect) and diffuse atrophy. In glaucoma, diffuse RNFL atrophy is thought to be more common than localized defects. 8 To date, several studies have focused on the performance of OCT in eyes with localized RNFL defects. 9 11 However, limited information is available regarding the ability of OCT for identifying diffuse RNFL atrophy. This is a critical issue in clinical practice because it is often difficult to discriminate normal eyes from eyes with diffuse atrophy. 8 Recently, our group suggested that OCT RNFL thickness parameters were well correlated with the degree of diffuse RNFL atrophy. 12 We also showed that Stratus OCT had an excellent ability to distinguish normal from diffuse RNFL atrophy. However, with its internal normative database, diagnostic accuracy of Stratus OCT to detect diffuse RNFL atrophy is still unknown. Because general ophthalmologists tend to rely on the classification based on its intrinsic normative database, 13 it is of clinical importance to investigate this issue. In this study, we investigated the diagnostic sensitivity and specificity of Stratus OCT with its internal normative database for the detection of diffuse RNFL atrophy. METHODS This study is a part of the Diffuse Atrophy Imaging Study, which enrolled glaucoma patients with diffuse RNFL atrophy and healthy subjects at the Glaucoma Clinic of Seoul National University Boramae Hospital. Details of the study design were published elsewhere. 12 This study was approved by the Institutional Review Board of the Seoul National University Boramae Hospital and conformed to the Declaration of Helsinki. Informed consent was obtained from the subjects after explanation of the nature and possible consequences of the study. In cases in which both eyes of a subject were eligible for the study, only one eye was randomly chosen for inclusion. Study Subjects Inclusion Criteria for the Diffuse Atrophy Imaging Study. All subjects had best-corrected visual acuity of 20/40 or better, spherical equivalent refractive error within 5.00 diopters Investigative Ophthalmology & Visual Science, August 2011, Vol. 52, No. 9 6074 Copyright 2011 The Association for Research in Vision and Ophthalmology, Inc.
IOVS, August 2011, Vol. 52, No. 9 Diagnostic Accuracy of OCT for Diffuse Atrophy 6075 (D) and astigmatism within 3.00 D, open anterior chamber angle, good quality of color disc photography and red-free RNFL photography, and reliable visual fields ( 33% false positives, 33% false negatives, and 20% fixation losses). All subjects underwent a complete ophthalmologic examination, including visual acuity, manifest refraction, intraocular pressure (IOP) measurements by Goldmann applanation tonometry, slit-lamp examination, gonioscopy, dilated fundus examination, color sequential stereo disc photography and red-free RNFL photography (TRC-50IX; Topcon, Tokyo, Japan), Swedish interactive thresholding algorithm (SITA) 30 2 perimetry (Humphrey Field Analyzer II; Carl Zeiss Meditec, Dublin, CA), and OCT (Stratus OCT; Carl Zeiss Meditec). All examinations were conducted within a 3-month period. Exclusion Criteria for the Diffuse Atrophy Imaging Study. Participants were excluded if they had a history of intraocular surgery other than uncomplicated cataract surgery. Also excluded were all subjects with nonglaucomatous secondary causes of elevated IOP (e.g., uveitis, trauma), other intraocular eye diseases, or other diseases affecting the visual field (e.g., diabetic retinopathy, retinal vein occlusion, ischemic optic neuropathy, pituitary lesions, and demyelinating diseases) or with diseases that could affect the peripapillary area where OCT measurements are obtained (e.g., large peripapillary atrophy, chorioretinal coloboma, and peripapillary staphyloma). Study Subjects and Controls. Diffuse RNFL atrophy eyes were defined as those having a generalized loss of RNFL visibility of the upper or lower retina without localized wedge-shaped RNFL defects regardless of their width. 12 To be eligible for the study, patients had to have glaucomatous optic neuropathy defined according to the following criteria: characteristic glaucomatous optic nerve head damage such as cup/disc asymmetry between fellow eyes of greater than 0.2, rim thinning, notching, or excavation and no neurologic disorder that could affect the optic nerve. Two glaucoma specialists (JWJ, SHK) assessed the stereo optic disc photographs in a masked fashion. Each observer estimated the horizontal and vertical cup-to-disc ratios on the basis of the contour of the cup using the color disc images. The mean value of both observers was used as the final grade. In addition, the presence of glaucomatous optic neuropathy was determined by a consensus agreement between both observers. Eyes with diffuse RNFL atrophy were divided into two groups: eyes with diffuse RNFL atrophy, accompanied by normal standard automated perimetry results (group 1, preperimetric diffuse RNFL atrophy group), and eyes with diffuse RNFL atrophy, accompanied by glaucomatous visual field defects in the corresponding hemifield location (group 2, perimetric diffuse RNFL atrophy group). Healthy control eyes had an IOP of 21 mm Hg with no history of increased IOP, an absence of glaucomatous disc appearance, no visible RNFL defect according to red-free RNFL photography, and a normal visual field by standard automated perimetry. Healthy control subjects who were age-matched with subjects of diffuse RNFL atrophy were selected for analysis. Diagnostic Tests Red-free RNFL Photography and Grading Methods for Diffuse RNFL Atrophy. Red-free RNFL photographs were acquired using a digital fundus camera after maximum pupil dilation. Fifty-degree views of the fundus, carefully focused on the retina using the built-in split-line focusing device, were obtained and reviewed on an LCD monitor. 10,11,14,15 Evaluation of RNFL photographs was completed using a semiquantitative method described by Quigley et al. 16 The brightness, texture, and covering of blood vessels by nerve fibers in the inferior and superior poles were assessed to provide a semiquantitative score for diffuse RNFL damage. In this study, each photograph was assessed by two graders, in a random order and masked fashion, who had no knowledge of the clinical information. Any disagreements were resolved through discussion, and, if necessary, a third grader was consulted. Based on these features, diffuse atrophy was divided into mild (D1), moderate (D2), and severe (D3) grades. A score of D0 indicated healthy nerve fiber, whereas a score of D3 indicated severe RNFL damage, with clearly visible blood vessels and no visible RNFL fibers. For each method, superior and inferior arcuate bundles were scored separately. Our previous study confirmed a substantial interobserver agreement in this assessment ( 0.760 and 0.777 for superior and inferior RNFL areas, respectively). 12 Optical Coherence Tomography. Stratus OCT was performed using a previously described technique. 3,9 11,14,15 Subjects were measured using the peripapillary fast RNFL scan of Stratus OCT, analyzed with software version 4.0. Only well-focused images, signal strength 6, and the presence of a centered circular ring around the optic disc were included in the analyses. If more than two OCT images were obtained, those of higher signal strength were selected for analysis. For securing centration, we selected the subjects with good fixation and used the repeat mode of the internal fixation technique constant for each scan. The following OCT parameters were used in the analysis: average, quadrant, clock-hour RNFL thicknesses, and temporal-superior-nasalinferior-temporal (TSNIT) thickness graphs. Right eye orientation was used for documentation of the quadrant and clock hour measurements. Twelve o clock corresponds to the superior region, 3 o clock corresponds to the nasal region, 6 o clock corresponds to the inferior region, and 9 o clock corresponds to the temporal region for both right and left eyes. 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. 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. Visual Field Testing. Visual field analysis was performed with the SITA standard of the perimeter (Humphrey Field Analyzer II 750; Carl Zeiss Meditec) using the central full-threshold program 30 2. Glaucomatous visual field loss was defined as the consistent presence of a cluster of three or more non-edge points on the pattern deviation plot with a probability of occurrence in 5% of the healthy population (P 5%), with one of these points having the probability of occurring in 1% of the healthy population (P 1%), a pattern standard deviation with P 5%, or a glaucoma hemifield test result outside normal limits. Visual field defects had to be repeatable on at least two consecutive tests at two separate visits with repeatable location of defect. Visual fields were evaluated for reliability and excluded if either the false-positive or the false-negative rate was 33% or the fixation loss was 20%. Statistical Analysis Sensitivity and specificity of the Stratus OCT parameters were tested by comparison with the internal normative database. Tested parameters were the global average RNFL thickness at the 5% and 1% levels, 1 quadrant abnormal at the 5% and 1% levels, 1 clock hour abnormal at the 5% and 1% levels, and any segment abnormal at the 5% and 1% levels in the TSNIT thickness graph. Segments abnormal at the 5% and 1% levels in the TSNIT thickness graph indicated those of the TSNIT thickness graph located below the yellow band (outside the 95% normal limit) and in the red band (outside the 99% normal limit). For the various OCT parameters, the likelihood ratios (LRs) and interval LRs were calculated. Several cutoff values were arbitrarily created for the interval LRs. The methods regarding the LRs have been described elsewhere in detail. 7,17,18 According to the classification suggested by Jaeschke et al., 18 LRs higher than 10 or lower than 0.1 would be associated with large effects on posttest probability, LRs from 5 to 10 or from 0.1 to 0.2 would be associated with moderate effects, LRs from 2 to 5 or from 0.2 to 0.5 would be
6076 Jeoung et al. IOVS, August 2011, Vol. 52, No. 9 TABLE 1. Characteristics of Study Subjects Normal Control (n 102) Diffuse Atrophy Group (n 102) P Age, y 63.2 12.3 63.1 12.0 0.953* Sex ratio, male/female 60:42 64:38 0.667 Refraction, diopters 0.92 2.23 0.76 2.94 0.696* IOP without medication, mm Hg 16.3 3.2 16.5 4.2 0.657* Humphrey C30 2 threshold visual field Mean deviation, db 0.28 1.78 8.27 8.55 0.001* Pattern standard deviation, db 1.90 0.50 6.71 4.68 0.001* OCT signal strength 6 9 (8.8%) 9 (8.8%) 7, 8 68 (66.7%) 62 (60.8%) 0.631 9, 10 25 (24.5%) 31 (30.4%) IOP, intraocular pressure; OCT, optical coherence tomography. * Comparison was performed using Student s t-test. Comparison was performed using chi-square test. associated with small effects, and LRs closer to 1 would be insignificant. The 95% CIs for LRs were calculated using the method by Simel et al. 19 All analyses were performed using statistical analysis software (SPSS for Windows, version 19.0; SPSS Inc., Chicago, IL). RESULTS Study Subjects We analyzed 102 eyes of 102 patients with diffuse RNFL atrophy and 102 healthy eyes of 102 age-matched subjects who fulfilled the inclusion and exclusion criteria. Details of the qualifying process were published elsewhere. 12 In brief, 431 eyes of 431 subjects (247 glaucoma patients and 184 healthy control subjects) were initially enrolled. In glaucoma subjects, 16 eyes with unacceptable OCT scans and 129 eyes with localized defect or ambiguous information in RNFL photographs were excluded. Of the control subjects, 102 eyes of 102 subjects who were age matched with subjects of diffuse RNFL atrophy were selected. Mean age was 63.1 12.0 years in the diffuse RNFL atrophy group and 63.2 12.3 years for normal group. Sex, refraction, and initial IOP were similar between the healthy subjects and those with diffuse RNFL atrophy. Mean deviation of white-on-white perimetry was 0.28 1.78 db for the healthy group and 8.27 8.55 db for diffuse atrophy group. In this study, 91.2% of the OCT images had a signal strength 7 (Table 1). Sensitivity, Specificity, and Likelihood Ratios for OCT Parameters In the diffuse atrophy group, 36 eyes showed superior RNFL atrophy and 32 eyes showed inferior RNFL atrophy, according to red-free fundus photography. Thirty-four eyes had diffuse RNFL atrophy at both the superior and the inferior areas. There were 136 areas of diffuse RNFL atrophy in total. Of these 102 eyes, 27 eyes and 75 eyes, respectively, were classified as having preperimetric diffuse RNFL atrophy (group 1) and perimetric diffuse RNFL atrophy (group 2). The overall sensitivity and specificity of Stratus OCT for the detection of diffuse RNFL atrophy are detailed in Table 2. Using a criterion of abnormal at the 5% level, the overall sensitivity of the Stratus OCT parameters ranged from 61.5% to 84.5%, and the overall specificity ranged from 90.2% to 99.0%. The highest sensitivity was yielded with the criterion of the TSNIT thickness graph abnormal at the 5% level (a sensitivity of 84.5% and a specificity of 90.2%), followed by that of the 1 clock hour abnormal at the 5% level (a sensitivity of 82.4% and a specificity of 90.2%). Table 3 presents the diagnostic accuracy of Stratus OCT for detecting diffuse RNFL atrophy with regard to its visual field status. For preperimetric diffuse RNFL atrophy, the sensitivity of Stratus OCT was relatively low, ranging from 12.9% to 54.8%, with a criterion of abnormal at the 5% level. For perimetric diffuse RNFL atrophy, however, the sensitivity of Stratus OCT ranged from 74.4% to 92.3%. Table 4 shows diagnostic performances of Stratus OCT for different grades of diffuse RNFL atrophy. The severity of RNFL damage had a significant influence on the sensitivity of Stratus OCT. For grade D1 atrophy, using a criterion of abnormal at the 5% level, the sensitivity of Stratus OCT ranged from 25.4% to 62.7%; however, for grade D2 atrophy, the sensitivity ranged from 74.0% to 98.0%. For grade D3 atrophy, the sensitivity of overall Stratus OCT parameters was 100%. Table 5 presents the diagnostic sensitivity and specificity of the various OCT RNFL parameters. A similar pattern was seen when the superior and inferior RNFL areas were analyzed separately. For grades D1, D2, and D3 atrophy, the superior quadrant had a sensitivity of 41.0%, 83.3%, and 100.0%, respectively, and the inferior quadrant had a sensitivity of 35.0%, 88.5%, and 100.0%, respectively. The LRs with their 95% CIs were calculated for the Stratus OCT parameters after comparison with the intrinsic TABLE 2. Overall Sensitivity and Specificity with 95% Confidence Intervals of Stratus OCT for the Detection of Diffuse Retinal Nerve Fiber Layer Atrophy Overall Sensitivity (%) Stratus OCT Overall Specificity (%) 1 Clock hour the 5% level 82.4 (75.3 88.2) 90.2 (82.7 95.2) the 1% level 61.5 (53.1 69.4) 98.0 (93.1 99.7) 1 Quadrant the 5% level 70.9 (62.9 78.1) 97.1 (91.6 99.4) the 1% level 54.1 (45.7 62.3) 100.0 (96.4 100.0) Average RNFL thickness the 5% level 61.5 (53.1 69.4) 99.0 (94.6 99.8) the 1% level 54.1 (45.7 62.3) 100.0 (96.4 100.0) TSNIT thickness graph the 5% level 84.5 (77.6 89.9) 90.2 (82.7 95.2) the 1% level 66.9 (58.7 74.4) 98.0 (93.1 99.7) OCT, optical coherence tomography; RNFL, retinal nerve fiber layer; TSNIT, temporal-superior-nasal-inferior-temporal.
IOVS, August 2011, Vol. 52, No. 9 Diagnostic Accuracy of OCT for Diffuse Atrophy 6077 TABLE 3. Diagnostic Performances of Stratus OCT for the Detection of Preperimetric and Perimetric Diffuse Retinal Nerve Fiber Layer Atrophy Diffuse Atrophy without Corresponding VF Defect Diffuse Atrophy with Corresponding VF Defect Sensitivity Specificity Sensitivity Specificity 1 Clock hour the 5% level 54.8 88.9 89.7 90.7 the 1% level 9.7 96.3 75.2 98.7 1 Quadrant the 5% level 32.3 96.3 81.2 97.3 the 1% level 3.2 100.0 67.5 100.0 Average RNFL thickness the 5% level 12.9 100.0 74.4 98.7 the 1% level 3.2 100.0 67.5 100.0 TSNIT thickness graph the 5% level 54.8 88.9 92.3 90.7 the 1% level 19.4 96.3 79.5 98.7 OCT, optical coherence tomography; VF, visual field; RNFL, retinal nerve fiber layer; TSNIT, temporal-superior-nasal-inferior-temporal. normative database (Table 6). The LRs of outside-normal limits ranged from 3.00 to infinity, and those of borderline results ranged from 1.50 to 21.00. The LRs of within-normal limits results ranged from 0.37 to 0.98. For the superior RNFL area, the highest LRs at the 5% level were obtained at the 12 and 11 o clock sectors. For the inferior RNFL area, the highest LRs at the 5% level were obtained at the 6 and 7 o clock sectors. Interval LRs with their 95% CIs were also evaluated for the Stratus OCT superior and inferior quadrant parameters (Table 7). Superior thickness 80 mor 120 m and inferior thickness 100 m or 140 m were associated with moderate to large effects. Other test results were associated with small or insignificant effects on posttest probabilities of disease. DISCUSSION This study was designed to evaluate the diagnostic performance of Stratus OCT with its internal normative database to detect diffuse RNFL atrophy. This study has found that Stratus OCT can detect diffuse RNFL atrophy with moderate sensitivity and high specificity. However, the diagnostic performance with a normative database is significantly influenced by the severity of glaucoma. In this study, the overall sensitivity of the Stratus OCT parameters ranged from 61.5% to 84.5%, using a criterion of abnormal at the 5% level. However, for the eyes with diffuse RNFL atrophy and corresponding visual field defect, the sensitivity of Stratus OCT increased, ranging from 74.4% to 92.3%. These results are similar to those of the earlier studies on the performance of OCT for detecting glaucoma with manifest visual field defects. Budenz et al. 20 reported that the sensitivity of Stratus OCT for perimetric glaucoma ranges from 84% to 89%, using a criterion of abnormal at the 5% level. A study performed by our group obtained a sensitivity of 85.9% and a specificity of 97.4% for detecting localized RNFL defects accompanied by corresponding visual field defects. 9 However, the direct comparisons are limited because the included subjects and the criteria for defining the sensitivity and specificity are different from those of the present study. The present study showed that the current Stratus OCT parameters are not sensitive enough to identify diffuse RNFL atrophy without corresponding visual field defects. The highest sensitivity was 54.8%, which was obtained from the parameters of 1 clock hours and TSNIT thickness graph. Our study showed a generally higher performance than that of the previous study, which reported the highest sensitivity of 40.8% for identifying the preperimetric localized RNFL defects. 11 For detecting early RNFL defects, the number of scan points, which affects the diagnostic ability of imaging devices, is an important factor. We speculate that more scan points would be located in the diffuse RNFL atrophy area than in the small localized defects, resulting in the relatively higher diagnostic ability for diffuse RNFL atrophy than localized defects. It is well known that the diagnostic performance of an imaging device is heavily dependent on the severity of TABLE 4. Degrees of Diffuse Retinal Nerve Fiber Layer Atrophy and Diagnostic Performances of Stratus OCT Grade D1 Atrophy Grade D2 Atrophy Grade D3 Atrophy Sensitivity Specificity Sensitivity Specificity Sensitivity Specificity 1 Clock hour the 5% level 57.6 92.5 98.0 90.5 100.0 92.3 the 1% level 20.3 99.0 80.0 100.0 100.0 99.0 1 Quadrant the 5% level 39.0 98.1 86.0 97.6 100.0 96.2 the 1% level 10.2 100.0 70.0 100.0 100.0 100.0 Average RNFL thickness the 5% level 25.4 99.0 74.0 100.0 100.0 100.0 the 1% level 15.3 100.0 64.0 100.0 100.0 100.0 TSNIT thickness graph the 5% level 62.7 92.5 98.0 90.5 100.0 92.3 the 1% level 28.8 99.0 86.0 100.0 100.0 99.0 OCT, optical coherence tomography; RNFL, retinal nerve fiber layer; TSNIT, temporal-superior-nasalinferior-temporal.
6078 Jeoung et al. IOVS, August 2011, Vol. 52, No. 9 TABLE 5. Diagnostic Sensitivity and Specificity of the Various OCT RNFL Thickness Parameters Superior Diffuse RNFL Atrophy* Grade D1 Atrophy (n 31) Grade D2 Atrophy (n 22) Grade D3 Atrophy (n 17) Sensitivity Specificity Sensitivity Specificity Sensitivity Specificity Superior quadrant 41.0 97.4 83.3 95.8 100.0 100.0 Clock hours 10 17.9 100.0 50.0 100.0 100.0 94.1 11 48.7 97.4 91.7 100.0 100.0 100.0 12 superior 12.8 100.0 58.3 95.8 100.0 100.0 1 17.9 100.0 33.3 95.8 64.7 94.1 2 15.4 97.4 33.3 95.8 47.1 94.1 Inferior Diffuse RNFL Atrophy* Grade D1 Atrophy (n 18) Grade D2 Atrophy (n 24) Grade D3 Atrophy (n 24) Sensitivity Specificity Sensitivity Specificity Sensitivity Specificity Inferior quadrant 35.0 100.0 88.5 96.2 100.0 100.0 Clock hours 4 0.0 94.7 7.7 92.3 18.2 95.5 5 0.0 94.7 42.3 96.2 54.5 100.0 6 inferior 15.0 100.0 84.6 100.0 100.0 100.0 7 55.0 100.0 96.2 100.0 100.0 95.5 8 15.0 94.7 30.8 100.0 72.7 100.0 OCT, optical coherence tomography; RNFL, retinal nerve fiber layer. * The sensitivity and specificity of OCT parameters were tested using the criterion of abnormal at the 5% level. glaucomatous damage in the subjects studied. Medeiros et al. 21 demonstrated that the severity of visual field loss has a significant influence on the sensitivity of the imaging devices. Kanamori et al. 22 compared OCT findings in normal eyes, eyes with ocular hypertension, eyes with suspected glaucoma, and eyes with glaucoma. In that study, the highest sensitivity for the glaucoma suspect group was 53% with a fixed specificity of 90%, whereas a sensitivity of 74% at a specificity of 91% was obtained in the early perimetric glaucoma group based on the OCT parameter of inferior quadrant. Nouri-Mahdavi et al. 23 reported better discriminating ability of OCT for the detection of early glaucoma than for suspected glaucoma (defined by disc changes with normal visual fields). Our results confirmed that the sensitivity TABLE 6. Likelihood Ratios and 95% Confidence Intervals of Stratus OCT Parameters for the Detection of Diffuse Retinal Nerve Fiber Layer Atrophy Likelihood Ratios (95% CIs) Within Normal Limits Borderline Outside Normal Limits Average 0.48 (0.39 0.58) 9.00 (1.16 69.75) Infinity (NA) Quadrant Superior 0.44 (0.35 0.54) 21.00 (2.88 153.21) Infinity (NA) Temporal 0.71 (0.62 0.81) 5.50 (1.25 24.20) Infinity (NA) Inferior 0.43 (0.34 0.53) 10.00 (1.30 76.70) Infinity (NA) Nasal 0.92 (0.84 1.00) 2.60 (0.96 7.03) Infinity (NA) Clock hours 12 superior 0.60 (0.51 0.71) 21.00 (2.88 153.21) Infinity (NA) 11 0.43 (0.34 0.53) 20.00 (2.74 146.25) Infinity (NA) 10 0.65 (0.57 0.76) 13.00 (1.73 97.55) Infinity (NA) 9 temporal 0.86 (0.79 0.94) 4.50 (1.00 20.32) Infinity (NA) 8 0.69 (0.61 0.79) 9.00 (1.16 69.75) Infinity (NA) 7 0.37 (0.28 0.47) 17.00 (2.31 125.37) Infinity (NA) 6 inferior 0.49 (0.40 0.60) 14.00 (1.88 104.50) 39.00 (5.46 278.51) 5 0.78 (0.70 0.87) 6.00 (1.38 26.14) Infinity (NA) 4 0.98 (0.92 1.04) 1.50 (0.44 5.16) Infinity (NA) 3 nasal 0.98 (0.93 1.03) 2.00 (0.37 10.68) Infinity (NA) 2 0.78 (0.69 0.89) 4.80 (1.91 12.09) 3.00 (0.32 28.36) 1 0.70 (0.61 0.80) 4.00 (1.16 13.75) Infinity (NA) OCT, optical coherence tomography; CI, confidence interval; NA, not applicable.
IOVS, August 2011, Vol. 52, No. 9 Diagnostic Accuracy of OCT for Diffuse Atrophy 6079 TABLE 7. Interval Likelihood Ratios and 95% CIs for the Detection of Diffuse Retinal Nerve Fiber Layer Atrophy RNFL Thickness ( m) Likelihood Ratios (95% CIs) Superior Quadrant Inferior Quadrant 80 Infinity (NA) Infinity (NA) 80 100 4.29 (1.97 9.31) 9.50 (2.27 39.74) 100 120 1.15 (0.74 1.81) 1.18 (0.72 1.94) 120 140 0.11 (0.04 0.26) 0.23 (0.13 0.42) 140 0.00 0.13 (0.05 0.35) RNFL, retinal nerve fiber layer; CI, confidence interval; NA, not applicable. of Stratus OCT tends to increase with increasing severity of RNFL damage, which corresponded well with the results of the previous studies. The LR represents the magnitude of change from a physician s initial suspicion of disease (pretest probability) to the likelihood of disease after the test (posttest probability). 18 The report of diagnostic LRs for comparison with normative databases is more clinically useful than sensitivities and specificities because LRs provide information that can be more readily incorporated into clinical practice. This study confirmed that the highest LRs were obtained at the 6 and 7 o clock sectors for inferior RNFL and the 11 and 12 o clock sectors for superior RNFL, suggesting that these sectors are the best parameters for the discrimination between healthy eyes and eyes with diffuse RNFL atrophy. However, our results showed that the LRs of within-normal limits results ranged from 0.37 to 0.98, indicating that these parameters were generally associated with small changes in probability. Depending on the pretest probability of disease and the clinical situation in which the test is used, however, even small changes in probability may be clinically relevant. 7 In the previous study, we evaluated quantitatively the degree of diffuse RNFL atrophy by using continuous RNFL thickness measurements, which showed that Stratus OCT had an excellent ability to distinguish normal from diffuse RNFL atrophy. 12 The present study has focused on the clinical usefulness of the imaging device. In clinical practice, general ophthalmologists are likely to rely on the classification based on the internal normative database. Our results showed that Stratus OCT with a normative database had relatively excellent diagnostic performance for diffuse atrophy with visual field defects. However, eyes with early glaucoma without visual field defects or eyes with mild diffuse atrophy can be missed if one relies entirely on the classification based on the normative database. On the other hand, our results showed relatively high specificity even in the early stage of glaucoma, suggesting that abnormal classification is likely to indicate the presence of glaucoma. RNFL thickness is variable in healthy subjects; an approximately two times difference is noted between the 5th percentile and the 95th percentile in a normative database. Because of intersubject variability, the OCT-measured RNFL thickness in patients with early glaucoma, even though it is in the decreasing state, can be classified as normal. This is a limitation of any classifier with a specific cutoff because it compares a patient s RNFL thickness value with a normative database instead of with the patient s previous healthy RNFL thickness. This inherent limitation might explain the relatively low sensitivity of OCT for the detection of early-stage glaucoma. The present study clearly demonstrates the diagnostic ability of Stratus OCT for detecting diffuse RNFL atrophy. Our results have confirmed that the Stratus OCT, with its normative database, can detect diffuse RNFL atrophy with moderate sensitivity and high specificity. Moreover, this technology has some advantages because it provides completely objective and quantitative data regarding peripapillary RNFL status. Because disease severity has a significant influence on the diagnostic performance of imaging devices, clinicians should interpret OCT with prudence, especially in the early stage of glaucoma with diffuse RNFL atrophy. References 1. Hoyt WF, Frisen L, Newman NM. Funduscopy of nerve fiber layer defects in glaucoma. Invest Ophthalmol. 1973;12:814 829. 2. Bowd C, Weinreb RN, Zangwill LM. Evaluating the optic disc and retinal nerve fiber layer in glaucoma, I: clinical examination and photographic methods. Semin Ophthalmol. 2000;15:194 205. 3. Huang D, Swanson EA, Lin CP, et al. Optical coherence tomography. Science. 1991;254:1178 1181. 4. Hee MR, Izatt JA, Swanson EA, et al. Optical coherence tomography of the human retina. Arch Ophthalmol. 1995;113:325 332. 5. Schuman JS, Pedut-Kloizman T, Hertzmark E, et al. Reproducibility of nerve fiber layer thickness measurements using optical coherence tomography. Ophthalmology. 1996;103:1889 1898. 6. Blumenthal EZ, Williams JM, Weinreb RN, Girkin CA, Berry CC, Zangwill LM. Reproducibility of nerve fiber layer thickness measurements by use of optical coherence tomography. Ophthalmology. 2000;107:2278 2282. 7. Medeiros FA, Zangwill LM, Bowd C, Weinreb RN. Comparison of the GDx VCC scanning laser polarimeter, HRT II confocal scanning laser ophthalmoscope, and stratus OCT optical coherence tomograph for the detection of glaucoma. Arch Ophthalmol. 2004;122: 827 837. 8. Tuulonen A, Airaksinen PJ. Initial glaucomatous optic disk and retinal nerve fiber layer abnormalities and their progression. Am J Ophthalmol. 1991;111:485 490. 9. Jeoung JW, Park KH, Kim TW, Khwarg SI, Kim DM. Diagnostic ability of optical coherence tomography with a normative database to detect localized retinal nerve fiber layer defects. Ophthalmology. 2005;112:2157 2163. 10. Hwang JM, Kim TW, Park KH, Kim DM, Kim H. Correlation between topographic profiles of localized retinal nerve fiber layer defects as determined by optical coherence tomography and red-free fundus photography. J Glaucoma. 2006;15:223 228. 11. Kim TW, Park UC, Park KH, Kim DM. Ability of Stratus OCT to identify localized retinal nerve fiber layer defects in patients with normal standard automated perimetry results. Invest Ophthalmol Vis Sci. 2007;48:1635 1641. 12. Jeoung JW, Kim SH, Park KH, Kim TW, Kim DM. Quantitative assessment of diffuse retinal nerve fiber layer atrophy using optical coherence tomography: diffuse atrophy imaging study. Ophthalmology. 2010;117:1946 1952. 13. Vessani RM, Moritz R, Batis L, Zagui RB, Bernardoni S, Susanna R. Comparison of quantitative imaging devices and subjective optic nerve head assessment by general ophthalmologists to differentiate normal from glaucomatous eyes. J Glaucoma. 2009;18:253 261. 14. Jeoung JW, Kim TW, Kang KB, Lee JJ, Park KH, Kim DM. Overlapping of retinal nerve fibers in the horizontal plane. Invest Ophthalmol Vis Sci. 2008;49:1753 1757. 15. Jeoung JW, Park KH, Kim JM, et al. Optic disc hemorrhage may be associated with retinal nerve fiber loss in otherwise normal eyes. Ophthalmology. 2008;115:2132 2140. 16. Quigley HA, Reacher M, Katz J, Strahlman E, Gilbert D, Scott R. Quantitative grading of nerve fiber layer photographs. Ophthalmology. 1993;100:1800 1807. 17. Radack KL, Rouan G, Hedges J. The likelihood ratio: an improved measure for reporting and evaluating diagnostic test results. Arch Pathol Lab Med. 1986;110:689 693.
6080 Jeoung et al. IOVS, August 2011, Vol. 52, No. 9 18. Jaeschke R, Guyatt GH, Sackett DL. Users guides to the medical literature, III: how to use an article about a diagnostic test. B. What are the results and will they help me in caring for my patients? The Evidence-Based Medicine Working Group. JAMA. 1994;271:703 707. 19. Simel DL, Samsa GP, Matchar DB. Likelihood ratios with confidence: sample size estimation for diagnostic test studies. J Clin Epidemiol. 1991;44:763 770. 20. Budenz DL, Michael A, Chang RT, McSoley J, Katz J. Sensitivity and specificity of the StratusOCT for perimetric glaucoma. Ophthalmology. 2005;112:3 9. 21. Medeiros FA, Zangwill LM, Bowd C, Sample PA, Weinreb RN. Influence of disease severity and optic disc size on the diagnostic performance of imaging instruments in glaucoma. Invest Ophthalmol Vis Sci. 2006;47:1008 1015. 22. Kanamori A, Nakamura M, Escano MF, Seya R, Maeda H, Negi A. Evaluation of the glaucomatous damage on retinal nerve fiber layer thickness measured by optical coherence tomography. Am J Ophthalmol. 2003;135:513 520. 23. Nouri-Mahdavi K, Hoffman D, Tannenbaum DP, Law SK, Caprioli J. Identifying early glaucoma with optical coherence tomography. Am J Ophthalmol. 2004;137:228 235.