Monitoring Glaucomatous Progression Using a Novel Heidelberg Retina Tomograph Event Analysis

Monitoring Glaucomatous Progression Using a Novel Heidelberg Retina Tomograph Event Analysis Tessa Fayers, BSc, MRCOphth, Nicholas G. Strouthidis, MD, MRCOphth, David F. Garway-Heath, MD, FRCOphth Purpose: To describe an event analysis (EA) for monitoring Heidelberg Retina Tomograph (HRT) progression and to establish specificity, detection rate, and agreement with visual field progression by application to longitudinal data. Design: Retrospective analysis of data from a randomized controlled trial. Participants: One hundred ninety-eight ocular hypertensive and 21 control subjects. Methods: Change criteria were derived from rim area (RA) repeatability coefficients for different levels of image quality. Event analysis 1 (EA1) through EA4 were applied to longitudinal series of HRT images acquired from the ocular hypertensive and the control cohorts: EA1 (change confirmed in 2 of 3 consecutive tests in 1 or more sector), EA2 (2 of 3 in 2 or more sectors), EA3 (3 of 3 in 1 or more sector), EA4 (3 of 3 in 2 or more sectors). Specificity (1 minus false positive results) was estimated by the proportions of progressing controls and significantly improving subjects. Progression rates were compared with Advanced Glaucoma Intervention Study (AGIS) visual field (VF) criteria, an HRT trend analysis, and a VF trend analysis, with specificity matched at 95%. Main Outcome Measures: Estimated specificity, progression rate, and agreement between progression techniques. Results: Specificity estimates were 76.2% to 88.1% (EA1), 94.1% to 95.2% (EA2), 92.2% to 95.2% (EA3), and 99.1% to 100% (EA4). Of ocular hypertensive (OHT) subjects, 45.4%, 28.3%, 26.3%, and 16.2% were identified as progressing by each strategy, respectively. With specificity at 95%, 12.1% of OHT subjects progressed by both EA2 and AGIS criteria, with a median time to progression of 3.2 and 3.6 years, respectively. By EA2 alone, 16.2% progressed, and 9.6% progressed by AGIS criteria alone. The RA trend analysis identified 12% of OHT subjects as progressing. Conclusions: The HRT EA represents a simple technique, taking into account image quality. In this cohort, it had a higher detection rate of progression, at 95% specificity, than RA trend analysis and the VF progression criteria. Ophthalmology 2007;114:1973–1980 © 2007 by the American Academy of Ophthalmology.

An important aspect in the management of patients with glaucoma, and in the design of prospective clinical trials in glaucoma, is the ability to monitor disease progression reliably. Glaucoma is defined as a progressive optic neuropathy Originally received: August 21, 2006. Final revision: January 17, 2007. Accepted: January 18, 2007. Available online: July 26, 2007. Manuscript no. 2006-936. From the Glaucoma Research Unit, Moorfields Eye Hospital, London, United Kingdom. Presented at: Association for Research in Vision and Ophthalmology Annual Meeting, April 2006, Fort Lauderdale, Florida. Supported by a Friends of Moorfields Research Fellowship, Moorfields Eye Hospital, and an unrestricted grant from Heidelberg Engineering, Heidelberg, Germany. Drs Strouthidis and Garway-Heath have received research funding from Heidelberg Engineering. Dr Garway-Heath is a consultant to Carl Zeiss Meditec. Correspondence to David F. Garway-Heath, MD, FRCOphth, Glaucoma Research Unit, Moorfields Eye Hospital, 162 City Road, London EC1V 2PD, United Kingdom. E-mail: [email protected]. © 2007 by the American Academy of Ophthalmology Published by Elsevier Inc.

with characteristic associated visual field (VF) losses.1 Progression may be measured by functional changes, such as deterioration in VF, and structural changes at the optic nerve head and retinal nerve fiber layer. Despite the long-standing availability of devices that can measure VF objectively and repeatably—automated perimeters, such as the Humphrey Field Analyzer (Humphrey Instruments, Carl Zeiss Meditec, Inc., Dublin, CA)—and more recently, image the optic nerve head, there is no consensus as to what is the best method for monitoring glaucomatous progression. The ability to discriminate true change, over and beyond measurement variability, is recognized as a central requirement of any progression technique. Progression techniques may be broadly classified as either event analyses (EAs) or trend analyses. A trend analysis identifies progression by monitoring the behavior of a parameter over time. Event analyses identify progression when a measurement exceeds a predetermined criterion for change (or an event); it is assumed that any change below the criterion represents measurement variability and that changes above the criterion represent true disease change. ISSN 0161-6420/07/$–see front matter doi:10.1016/j.ophtha.2007.01.035

1973

Ophthalmology Volume 114, Number 11, November 2007 Because EAs generate a binary end point—progression or no progression—they are the VF progression technique of choice in prospective clinical trials.2– 4 In terms of monitoring structure, Heidelberg Retina Tomograph (HRT; Heidelberg Engineering GmbH, Heidelberg, Germany) rim area (RA) progression has been assessed using EAs.5–7 Rim area has been identified as a repeatable and reliable parameter in test–retest studies, with a meaningful clinical corollary; it therefore represents a suitable metric for measuring disease progression.8 –12 A robust estimate of RA measurement precision is necessary to define a criterion for glaucomatous change. The British Standards Institution defines measurement precision according to the repeatability coefficient (RC), whereby 95% of intertest differences are expected to lie within this range.13 The authors’ group recently published RC values for global RA at different levels of HRT image quality, as measured by mean pixel height standard deviation (MPHSD).11 As a strategy for monitoring glaucomatous progression, it is possible that global RA may not identify focal RA changes. A more suitable approach therefore would be to estimate RC values for RA sectors. It should be noted that localized RA variability within individual disc sectors is likely to exceed the global rim area variability. However, each sector’s change criterion value reflects this such that greater change is required to meet progression criteria for noisier sectors. In this study, a novel HRT EA was developed using RC values derived from interobserver and intervisit sectoral RA measurements in a repeatability study.10,11 The new technique was applied to longitudinal HRT data collected from ocular hypertension (OHT) patients and control subjects; specificity and progression rates were estimated. The HRT EA also was compared with an established VF EA, an RA trend analysis, and a VF trend analysis, applied to the same longitudinal data set.

Patients and Methods Defining Criteria for Change 10

An HRT test–retest study has been described in detail elsewhere. In summary, 43 eyes with OHT and 31 with primary open-angle glaucoma were included in the protocol. Subjects had no history of intraocular surgery and had previous experience of optic nerve head imaging with the HRT. A single eye from each subject was imaged by 2 experienced observers using the classic HRT on the same day (with 1 observer imaging the subject twice), and on a different day within 6 weeks of the first test. Single topographies acquired using classic HRT were imported into the HRT II operating software, Heidelberg Eye Explorer (version 1.7.0; Heidelberg Engineering GmbH) with which mean topographies were generated and analyzed. The 320-␮m reference plane was adopted because this was shown to result in less RA variability.9,11 Explorer defines 6 RA sectors: superotemporal, inferotemporal, temporal, superonasal, inferonasal, and nasal. To simulate the situation likely to be encountered in a clinical setting, the RC values for interobserver and intervisit sectoral RA were calculated. A baseline reference sectoral RA value (observation 1) was generated in 1 of 2 ways: (1) from a single mean topography or (2) from the mean of 2 mean topographies acquired on the same day by the same observer. The follow-up sectoral RA (observation 2) was

1974

acquired from a mean topography obtained on a subsequent day by a different observer. The RC was calculated as: RC ⫽ 2 ⴱ

冉冑 兺

(baseline ⫺ observation N)2 n(observations)

冊

,

where baseline ⫽ sectoral RA derived from either (1) a single baseline mean topography or (2) the mean of 2 mean topographies acquired by the same observer on the same day; observation N is the follow-up image sectoral RA, and n ⫽ 2. Subject eyes were classified according to image quality: good quality (mean MPHSD, ⬍21), medium quality (mean MPHSD, 21–35), and poor quality (mean MPHSD, ⬎35). The RC values were calculated for each of the 6 HRT sectors at each level of image quality.

Defining the Novel Event Analysis The RC values generated equate to estimates of interobserver and intervisit sectoral RA repeatability, whereby 95% of differences between repeated sectoral RA measurements were within the RC value. An intertest difference that exceeds the RC threshold is suggestive of change resulting from causes other than measurement error, such as genuine disease progression. Calling progression when the RC is exceeded on just one occasion is liable to result in a high false-positive rate over a long series of images, primarily because the change analysis is applied repeatedly over time (there is a 5% chance of erroneously identifying change at each visit). Schulzer et al14 and Schulzer15 indicated that the false-positive rate, and therefore specificity, may be increased by including additional confirmatory tests or by increasing the magnitude of the criterion for change. Because the RC values are predefined according to image quality, the criterion for change identification may be raised by increasing the number of disc sectors required to satisfy the progression criteria and by requiring change confirmation. The requirement for both spatial and temporal criteria to confirm progression has parallels with Glaucoma Progression Analysis, the VF progression software native to the Humphrey perimeter. Glaucoma Progression Analysis confirms progression when deterioration beyond the fifth percentile of test– retest variability is observed at 3 test locations on 3 consecutive tests.12 The following EA strategies were tested in this study: 1. Two of 3 criteria in ⱖ1 disc sectors (EA1). Progression is suggested if: follow-up ⫺ baseline sector RA (sector RA difference) ⬎ sector RC for the level of image quality, with image quality defined as the mean MPHSD ([baseline ⫹ follow-up MPHSD]/2). Progression is confirmed if: sector RA difference ⬎ RC in at least 1 of the next 2 consecutive follow-up images, in 1 sector or more. 2. Two of 3 criteria in ⱖ2 disc sectors (EA2). As for the previous strategy, except change is required in ⱖ2 disc sectors. 3. Three of 3 criteria in ⱖ1 disc sectors (EA3). Progression is suggested if: follow-up ⫺ baseline sector RA ⬎ sector RC for the level of image quality. Progression is confirmed if: sector RA difference ⬎ RC in the next 2 consecutive follow-up images, in 1 sector or more. 4. Three of 3 criteria in ⱖ2 disc sectors (EA4). As for the previous strategy, except change is required in ⱖ2 disc sectors. Significant improvement for all 4 strategies is identified if an increase in sectoral RA occurs from baseline to follow-up image and the difference exceeds the RC value of that sector for the level of image quality.

Fayers et al 䡠 Novel HRT Event Analysis Testing the Novel Event Analysis Using Longitudinal Data Strategies EA1 through EA4 were applied to longitudinal HRT mean topographies acquired using the classic HRT derivative from 2 cohorts: an OHT cohort of 198 subjects and a control cohort of 21 subjects, followed prospectively with regular HRT and VF testing from 1994 through 2001. These 2 cohorts were recruited originally as part of a betaxolol versus placebo study and have been described in detail elsewhere.16,17 In brief, OHT was defined as an intraocular pressure (IOP) of more than 22 mmHg and less than 35 mmHg on 2 or more occasions within a 2-week period and a baseline mean Advanced Glaucoma Intervention Study (AGIS) VF score of 0 (Humphrey Field Analyzer, full-threshold 24-2 program).2 Control subjects were recruited from senior citizens groups or were the spouses or friends of subjects in the OHT cohort; they were not attending the eye clinic as patients and were not seeking care nor undergoing check-ups.5 Controls had a baseline IOP of less than 21 mmHg, normal baseline VF test results (same criteria as in the OHT group), and were excluded if there was a family history of glaucoma or any coexistent ocular or neurologic pathologic features. In the current study, the same eye was selected for analysis as had been randomized in the original study. The OHT eye randomization was based on risk to glaucomatous conversion, classified according to pattern electroretinogram results, IOP, and cup-to-disc area ratio at the time of recruitment16; control eyes were selected by simple randomization. The baseline characteristics and demographics of both cohorts are summarized in Table 1. The HRT mean topographies were generated and analyzed using Heidelberg Eye Explorer (version 1.7.0; Heidelberg Engineering GmbH). The 320-␮m reference plane was used for all analyses. Contour lines were drawn by a single observer (NGS) onto the baseline mean topographies and were exported automatically to the subsequent images. A manual alignment facility was used to correct contour line position if the automatically placed contour line was misplaced or if there was a magnification change.11 A minimum of 5 HRT mean topographies were available for each subject, with images of all qualities selected for analysis except where satisfactory contour line alignment could not be achieved. In total, 8 mean topographies were excluded from the study, either as a result of double imaging or if the image was so grainy as to prevent adequate visualization of Elschnig’s ring. For each longitudinal series, the baseline image was taken as the first mean topography available for each subject; the baseline visual field was taken as the test that coincided with the baseline HRT mean topography. When HRT testing was introduced to the bet-

axolol versus placebo study, imaging took place at yearly intervals. Therefore, it was not possible to construct a mean baseline image from 2 mean topography images. The criteria for change selected therefore were those calculated using only a single baseline mean topography. Specificity (1 minus false-positive results) was estimated for EA1 through EA4 using 2 proxy measures: the number of control subjects (of 21) progressing and the number of subjects (of 219) demonstrating significant improvement. The number of OHT subjects identified as progressing (positive hit rate) also was compared for the 4 strategies.

Comparison with Other Progression Strategies As part of the ongoing examination of subjects recruited to the betaxolol versus placebo study, subjects underwent full-threshold 24-2 Humphrey VF testing at approximately 4-month intervals. The AGIS VF scoring was performed for each subject’s VF series.2 Visual field progression was identified if the AGIS VF score increased from 0 to 1 or more and was reproducible in 3 consecutive VF tests in the same region of the VF. A glaucoma expert independently confirmed the VF series that were identified as progressing. Previous studies have estimated the specificity of the AGIS VF strategy, an EA, at between 91% and 100%.18 –20 Agreement with regard to progression status was assessed for the 2 event analyses (HRT and VF) using the HRT technique with an estimated specificity closely approximating that of the AGIS VF criteria. Time to identified progression also was compared; this was taken as the date of the final confirmatory test. An HRT trend analysis, based on linear regression of sector RA and time, recently was described by the authors’ group.17 A high stringency HRT progression strategy achieved specificity estimates of 95.2% to 98.2% using the same longitudinal data as was used in the current study. The 3-omitting pointwise linear regression VF criterion achieved the same specificity estimates in the same subjects. Agreement with regard to progression status was assessed for these 2 trend analyses, the AGIS VF strategy and the novel HRT EA. To validate the comparison between the 4 progression criteria, the estimated specificity for each needed to be matched; an HRT EA strategy with specificity approaching 95% therefore was selected. The study adhered to the tenets of the Declaration of Helsinki and had local ethical committee approval, in addition to subjects’ informed consent. All statistical analyses were performed using Medcalc version 7.4.2.0 (Medcalc Software, Mariakerke, Belgium).

Table 1. Demographic Details and Baseline Characteristics of Ocular Hypertensive and Control Subjects Analyzed in the Study Characteristics

Ocular Hypertension Group

Control Group

No. of subjects Male-to-female ratio Laterality (right-to-left) Median age (range), yrs Median follow-up (range), yrs Median no. of HRT examinations Median no. of visual field examinations Median baseline mean defect (range), dB Median baseline global rim area (range), mm2 Median image quality throughout study (range), MPHSD

198 105:93 95:103 60 (32–79) 6.0 (2.3–7.2) 10 (5–16) 17 (5–33) ⫹0.1 (⫺2.7 to ⫹3.0) 1.24 (0.63–2.31) 20 (7–186)

21 14:7 11:10 65 (41–77) 5.3 (3.1–6.8) 9 (8–11) 9 (7–14) ⫹0.1 (⫺2.4 to ⫹2.6) 1.35 (0.86–2.51) 23 (9–80)

dB ⫽ decibel; HRT ⫽ Heidelberg Retina Tomography; MPHSD ⫽ mean pixel height standard deviation.

1975

Ophthalmology Volume 114, Number 11, November 2007 Table 2. Repeatability Coefficients for Interobserver, Intervisit Sectoral Rim Area Measurements Rim Area Repeatability Coefficients

Temporal Sector

Nasal Sector

Good image quality (MPHSD ⬍21), mm2 Medium image quality (MPHSD 21–35), mm2 Poor image quality (MPHSD ⬎35), mm2

0.050 (0.054) 0.058 (0.042) 0.109 (0.096)

0.014 (0.016) 0.031 (0.030) 0.068 (0.079)

Superotemporal Inferotemporal Sector Sector 0.012 (0.013) 0.026 (0.022) 0.057 (0.047)

0.026 (0.028) 0.026 (0.023) 0.046 (0.042)

Superonasal Sector

Inferonasal Sector

0.008 (0.007) 0.022 (0.021) 0.049 (0.042)

0.007 (0.006) 0.015 (0.014) 0.029 (0.029)

MPHSD ⫽ mean pixel height standard deviation. Baseline rim area has been calculated using a single image to derive the baseline rim area. Figures in parentheses are for repeatability coefficients calculated using 2 images acquired on the same day by the same observer. Repeatability coefficients are stratified according to Heidelberg Retina Tomograph (HRT) image quality, measured by mean pixel height standard deviation.

Results The criteria for change, identified by calculating interobserver and intervisit RA repeatability coefficients using a single baseline mean topography, are summarized in Table 2. The figures in parentheses represent RC values calculated using a baseline constructed from the mean of 2 HRT mean topographies acquired by the same observer at the same visit. In this category, the RC values for good quality images were higher than those of medium quality images in the temporal and the inferotemporal disc sectors. For the purposes of the event analyses, the RC value for the good quality images (0.054 in the temporal sector and 0.028 in the inferotemporal sector) may be adopted as the criterion for change for both good- and medium-quality images in that sector. There was no statistically significant difference between RC values generated from a single baseline mean topography and an average of 2 mean topographies (P ⫽ 0.76, F test). In general, however, single mean topography baseline RA values generated slightly higher repeatability coefficients for medium- and poor-quality images and slightly lower repeatability coefficients for good-quality images compared with the 2 mean topographies baseline. The estimated specificities for EA1 through EA4 are summarized in Table 3. Event analyses 1 through 4 identified 90 (45.4%), 56 (28.3%), 52 (26.3%), and 32 (16.2%) OHT subjects, respectively, as progressing at the end of the study period. Event analysis 2 achieved a specificity estimate of approximately 95%, compared with approximately 99% for EA4. For the comparison with AGIS VF criteria, HRT trend analysis, and pointwise linear regression to be valid, a similar level of specificity for each technique is required. It is for this reason that EA2 was selected for the comparisons. The choice of level of specificity that is required of a progression criterion largely is arbitrary, and a false-positive rate of 5% (95% specificity) generally is acceptable. What is most important is that our technique allows the false-positive rate to be estimated, and this aids clinical interpretation.

Comparing EA2 with AGIS VF criteria, 24 OHT subjects were identified as progressing by both disc and by VF. A further 32 OHT subjects progressed by HRT alone and 19 by VF alone (Fig 1). The age, mean deviation (MD), global RA at baseline, and MPHSD throughout the series for subjects progressing by HRT alone or by VF alone are summarized in Table 4. There was a significant difference in baseline MD (P ⫽ 0.02, unpaired t test) and MPHSD throughout the series (P ⫽ 0.0001, unpaired t test) between the 2 groups, with those progressing by HRT alone having better image quality and more positive MD. The median time to detection of progression in the OHT group was 3.2 years (range, 1.2–5.5 years) using EA2 and 3.6 years (range, 1.0 –7.2 years) using AGIS VF criteria. The time to identification of progression in the OHT cohort, using the 2 strategies, was compared in the Kaplan-Meier survival curve shown in Figure 2. Comparing EA2 with the high stringency RA trend analysis (slope more than 1% of baseline sector RA/year; P⬍0.001 for low variability series and P⬍0.0001 for high variability series) within the OHT cohort, 19 subjects (9.6%) had progressed by both strategies at the end of the study period. A further 37 subjects (18.7%) progressed by EA2 alone, and a further 5 subjects (2.5%) progressed by HRT trend analysis alone. Comparing AGIS VF criteria and 3-omitting pointwise linear regression in the OHT cohort, 26 subjects progressed by both strategies, 17 subjects progressed by AGIS criteria alone, and 11 subjects progressed by 3-omitting pointwise linear regression alone. Five subjects progressed by all 4 techniques. The number of OHT subjects identified as progressing by each technique at the end of the study is compared in Figure 3.

Discussion The ideal attributes for any strategy implemented for the detection of glaucomatous progression are high sensitivity,

Table 3. Estimation of Specificity for 4 Novel Heidelberg Retina Tomograph Event Analysis Strategies Event Analysis

Event Analysis 1*

Event Analysis 2†

Event Analysis 3‡

Event Analysis 4§

Subjects without significant improvement 95% confidence limits Controls without progression 95% confidence limits Specificity estimate

193/219 (88.1%) 82.9%–91.9% 16/21 (76.2%) 52.4%–90.9% 76.2%–88.1%

206/219 (94.1%) 89.8%–96.7% 20/21 (95.2%) 74.1%–99.7% 94.1%–95.2%

202/219 (92.2%) 87.6%–95.3% 20/21 (95.2%) 74.1%–99.7% 92.2%–95.2%

217/219 (99.1%) 96.4%–99.8% 21/21 (100%) 80.7%–100% 99.1%–100%

HRT ⫽ Heidelberg Retina Tomograph. *Progression suggested in a minimum of 1 HRT disc sector and confirmed by at least 1 of the next 2 consecutive HRT tests. † Progression suggested in a minimum of 2 HRT disc sectors and confirmed by at least 1 of the next 2 consecutive HRT tests. ‡ Progression suggested in a minimum of 1 HRT disc sector and confirmed by the next 2 consecutive HRT tests. § Progression suggested in a minimum of 2 HRT disc sectors and confirmed by the next 2 consecutive HRT tests.

1976

Fayers et al 䡠 Novel HRT Event Analysis

Figure 1. Venn diagram comparing Heidelberg Retina Tomograph (HRT) progression (2 of 3, in 2 or more sectors, strategy) and Advanced Glaucoma Intervention Study visual field progression in the longitudinal ocular hypertensive cohort, expressed as a percentage of subjects. Estimated specificity is 95%.

high specificity, the ability to detect progression early, and resistance to high levels of measurement variability. Kamal et al5,6 estimated normal RA variability by deriving 95% confidence limits for change from RA differences measured in sequential HRT images acquired from control subjects. In terms of defining a criterion for change, this method does not take into account the variation in interindividual measurement error (images from some eyes are more variable than others). Tan and Hitchings7,21 avoided this problem by deriving limits for RA change (VARLIM) for each eye from the single topography images acquired at each visit in a longitudinal series. Using the same 2 of 3 confirmatory strategy used in EA2, Tan and Hitchings7 reported a sensitivity (identified HRT progression in those meeting the AGIS VF criteria for progression) of 85% and a specificity of 95%. A comparable figure in this study is 56%. It should be noted that an experimental reference plane was adopted in the application of Tan and Hitchings’s7 progression strategy and not the 320-␮m reference plane that was used in the HRT progression strategies used in the current study. A major drawback of the VARLIM strategy is that single topography measures are not readily available on current Explorer software (both for the HRT 2 and, more recently, for the HRT 3); therefore, it is not yet possible to estimate the variability limits for RA change in this context. The current study estimates RA measurement error from repeated acquisitions from a cohort of OHT subjects and a cohort of glaucoma subjects. The recruitment of subjects for the test–retest study was enriched artificially by including the eye with greater lenticular opacity compared with its fellow.10 This was done to ensure a wide degree of measurement variability. The subjects recruited also had a wide range of disease severity and included those with preperi-

metric or early glaucoma and advanced glaucoma and OHT subjects with no glaucomatous features. An advantage of the test–retest data set is that it was repeated using the HRT 2. Criteria for change therefore are accessible for images acquired using the newer device. Similarly, repeatability coefficients can be calculated for HRT 2 acquisitions analyzed using the HRT 3 software in which a new image alignment algorithm operates.22 A potential weakness is that the test– retest data may not be applicable equally to other populations with different racial and physiologic characteristics. A unique element in the novel EA is the stratification according to image quality. Although it is not clear whether MPHSD truly equates to a subjective appreciation of image quality, it is a measure of variability (of pixel height) and has been shown to influence RA variability.10,23 Applying differing criteria for change according to image quality (MPHSD) should result in minimization of false-positive progression. Estimated specificity in this study was increased by 2 techniques: first, by increasing the number of confirmatory images (3 of 3), and second, by requiring change in more than 1 disc sector. A potential weakness of the latter strategy is that it may result in a diminution of spatial resolution. The temporal and nasal HRT sectors have the highest RC values, regardless of image quality. Fluctuation in topographic height has more impact on RA variability in the temporal portion of the rim, resulting in the greater degree of noise in this sector because the temporal neuroretinal rim has a gentler contour than other disc sectors. In less steep portions of the disc, small changes in topographic height results in greater changes in RA magnitude. This effect was observed despite the use of the more stable 320-␮m reference plane. This may be particularly prevalent in discs with oblique insertion into the globe. Greater RA variability at the nasal sector is likely to be a manifestation of the distribution of larger blood vessels in that sector. The position and size of blood vessels will vary according to the patient’s pulse, blood pressure, and fixation— factors unlikely to be constant between image acquisitions. Because a greater proportion of larger blood vessels are located in the nasal sector, RA measurements are likely to be more variable than in sectors with fewer blood vessels. The authors’ group previously applied the novel EA using global RA and a 2-of-3 change criterion (Invest Ophthalmol Vis Sci 47:e-abstract 898, 2006). Using the same longitudinal HRT data as was used in the current study, a specificity range of 95.2% to 98.6% was estimated. The positive detection rate was 14.2%, which is approximately

Table 4. Characteristics of Subjects Progressing by Optic Disc Alone and by Visual Field Alone Ocular Hypertensive Subjects’ Characteristics Median Median Median Median

Heidelberg Retina Tomograph Event Analysis 2

Visual Field Alone (Advanced Glaucoma Intervention Study)

P Value*

63.0 (45.4–78.9) 18 (8–186) 1.1 (0.8–1.6) 0.4 (⫺1.9 to ⫹2.9)

69.9 (41.0–74.1) 30 (9–114) 1.2 (0.6–2.1) ⫺0.2 (⫺1.5 to ⫹2.4)

0.23 ⬍0.0001 0.17 0.02

baseline age (range), yrs image quality throughout study (range), MPHSD baseline global rim area (range), mm2 baseline mean deviation (range), dB

dB ⫽ decibel; MPHSD ⫽ mean pixel height standard deviation. *Unpaired t test.

1977

Ophthalmology Volume 114, Number 11, November 2007

Figure 2. Kaplan-Meier survival curve comparing time to detection of progression in the longitudinal ocular hypertension (OHT) cohort using the Heidelberg Retina Tomograph (HRT) Event Analysis 2 (2 of 3 in 2 or more sectors) and Advanced Glaucoma Intervention Study (AGIS) visual field (VF) criteria.

half that of EA2, which has a similar level of estimated specificity. This comparison indicates that the adoption of local indices to identify change results in a higher detection rate than a global index at the same level of specificity. Because the estimated specificity is high, it is reasonable to assume that the HRT progressors identified by both local and global techniques are likely to be genuine. In comparing structural and functional progression, the choice of VF EA was pragmatic. In the absence of a method

Figure 3. Venn diagram comparing the number of ocular hypertensive subjects identified as progressing at the end of the longitudinal study period by 4 progression strategies. Heidelberg Retina Tomograph (HRT) event analysis 2 ⫽ 2 of 3 in ⱖ2 sectors; HRT trend analysis ⫽ slope ⬎ 1% of baseline sector rim area/year; n ⫽ no. of subjects identified as progressing using each strategy; VF trend analysis ⫽ 3-omitting pointwise linear regression of sensitivity/time; visual field (VF) event analysis ⫽ Advanced Glaucoma Intervention Study VF criteria. The specificity for the 4 techniques is matched at approximately 95%. P⬍0.001 for low-variability series; P⬍0.0001 for high-variability series.

1978

for external verification, there is no optimal single VF progression strategy. The AGIS VF scoring was part of the protocol of the betaxolol versus placebo study, and this was continued after the completion of the study in 1998, until the conclusion of the current study. Shortcomings of the AGIS VF criteria compared with other VF event analyses are emphasized in the literature. These include high rates of reversal—a lack of maintained progression status24,25—and being excessively conservative, with similar specificity but lower detection rates than the Collaborative Initial Glaucoma Treatment Study VF criteria.20 The VF EA used in the Early Manifest Glaucoma Trial examines changes in the pattern deviation.4 This technique is used in the Glaucoma Progression Analysis—the native progression algorithm for the Humphrey Field Analyzer 2. This technique may screen out field loss resulting from cataract; pattern deviation strategies have relatively poor agreement with total deviationbased strategies (such as the Collaborative Initial Glaucoma Treatment Study VF scoring system).26 A recent report has suggested, however, that pattern deviation systems may underestimate VF progression in the absence of increasing media opacity.12 At present, it is not possible to apply the commercially available Early Manifest Glaucoma Trial strategy to retrospective full-threshold VF series as used in this study. A considerable proportion of subjects (79%) identified as progressing by the HRT trend analysis also were identified by the novel HRT EA. Given similar levels of specificity, the EA has a much higher detection rate, with an additional 19 subjects progressing by HRT EA alone. Fifty-one percent of subjects identified as progressing by VF trend anal-

Fayers et al 䡠 Novel HRT Event Analysis ysis also progressed by HRT EA, compared with 56% of subjects identified by the AGIS VF criteria. Although there is a widely held view that structural changes are detected before functional changes using current testing techniques,27 there was no significant difference in the time to detection of progression by the different event analyses. This may reflect subjects commencing the study at different stages of the disease process; this is particularly pertinent because optic disc appearance was not taken into consideration at the time of recruitment. The results of the study show that eyes progressing by VF criteria alone had a lower MD, on average, at baseline. The new HRT EA represents a simple technique that performs at least as well, if not better, in terms of detection rate as other established techniques. Currently, the EA may be applied by exporting HRT data into a spreadsheet function; however, it is hoped that the EA may be incorporated into future iterations of the Explorer software. A potential weakness of the use of repeatability coefficients to define threshold for change is that the HRT test–retest RA noise is not normally distributed, but is best characterized by a hyperbolic distribution.28 However, for the purposes of this study, the use of empirical estimates of test specificity, as well as temporal and spatial confirmatory steps, should counteract any imprecision likely to have resulted from the distribution of RA variability. In the present study, only a single mean topography was available at baseline. Although there is no statistically significant difference, the RCs generated in this study suggest that the acquisition of 2 mean topographies to generate a baseline may be of particular benefit with medium- and poor-quality images; a single baseline mean topography should suffice for good-quality images. This study adds to the weight of evidence suggesting that there is poor agreement between structural and functional measurements of progression despite similar high levels of specificity and regardless of stage of disease.17,29 The methods for detecting progression in this study seem to identify different subgroups with OHT progressing by structure or function, with very little overlap. The discrepancy may be the result of functional changes occurring without structural changes (such as IOP-dependent ganglion cell dysfunction) or vice versa (such as lamina bowing or astrocyte loss). It is also possible that a degree of the disagreement may be the result of differences in measurement variability between the 2 methods in individual patients. The present results support this hypothesis, with eyes progressing by HRT criteria alone having better image quality than those progressing by VF alone. It is not possible to be certain that there is an ordered structure– function relationship in glaucoma until the issue of test variability has been resolved fully. Future developments in the identification of glaucomatous progression will need to limit the influence of measurement variability by improving the acquisition and the processing of both images and visual fields.

References 1. Gupta N, Weinreb RN. New definitions of glaucoma. Curr Opin Ophthalmol 1997;8:38 – 41.

2. Advanced Glaucoma Intervention Study. 2. Visual field test scoring and reliability. Ophthalmology 1994;101:1445–55. 3. Musch DC, Lichter PR, Guire KE, et al. The Collaborative Initial Glaucoma Treatment Study: study design, methods, and baseline characteristics of enrolled patients. Ophthalmology 1999;106:653– 62. 4. Heijl A, Leske MC, Bengtsson B, et al. Measuring visual field progression in the Early Manifest Glaucoma Trial. Acta Ophthalmol Scand 2003;81:286 –93. 5. Kamal DS, Viswanathan AC, Garway-Heath DF, et al. Detection of optic disc change with the Heidelberg retina tomograph before confirmed visual field change in ocular hypertensives converting to early glaucoma. Br J Ophthalmol 1999;83: 290 – 4. 6. Kamal DS, Garway-Heath DF, Hitchings RA, Fitzke FW. Use of sequential Heidelberg retina tomograph images to identify changes at the optic disc in ocular hypertensive patients at risk of developing glaucoma. Br J Ophthalmol 2000;84:993– 8. 7. Tan JC, Hitchings RA. Approach for identifying glaucomatous optic nerve progression by scanning laser tomography. Invest Ophthalmol Vis Sci 2003;44:2621– 6. 8. Tan JC, Garway-Heath DF, Hitchings RA. Variability across the optic nerve head in scanning laser tomography. Br J Ophthalmol 2003;87:557–9. 9. Tan JC, Garway-Heath DF, Fitzke FW, Hitchings RA. Reasons for rim area variability in scanning laser tomography. Invest Ophthalmol Vis Sci 2003;44:1126 –31. 10. Strouthidis NG, White ET, Owen VM, et al. Factors affecting the test-retest variability of the Heidelberg retina tomograph and Heidelberg retina tomograph-II measurements. Br J Ophthalmol 2005;89:1427–32. 11. Strouthidis NG, White ET, Owen VM, et al. Improving the repeatability of Heidelberg retina tomograph and Heidelberg retina tomograph II rim area measurements. Br J Ophthalmol 2005;89:1433–7. 12. Artes PH, Nicolela MT, LeBlanc RP, Chauhan BC. Visual field progression in glaucoma: total versus pattern deviation analyses. Invest Ophthalmol Vis Sci 2005;46:4600 – 6. 13. Precision of Test Methods 1. Guide for the Determination and Reproducibility for a Standard Test Method. London: BSI; 1979. BS 5497. 14. Schulzer M, Anderson DR, Drance SM. Sensitivity and specificity of a diagnostic test determined by repeated observations in the absence of an external standard. J Clin Epidemiol 1991;44:1167–79. 15. Schulzer M. Errors in the diagnosis of visual field progression in normal-tension glaucoma. Ophthalmology 1994;101:1589 – 94, discussion 1595. 16. Kamal D, Garway-Heath D, Ruben S, et al. Results of the betaxolol versus placebo treatment trial in ocular hypertension. Graefes Arch Clin Exp Ophthalmol 2003;241:196 –203. 17. Strouthidis NG, Scott A, Peter NM, Garway-Heath DF. Optic disc and visual field progression in ocular hypertensive subjects: detection rates, specificity and agreement. Invest Ophthalmol Vis Sci 2006;47:2904 –10. 18. Zahari M, Mukesh BN, Rait JL, et al. Progression of visual field loss in open angle glaucoma in the Melbourne Visual Impairment Project. Clin Experiment Ophthalmol 2006;34: 20 – 6. 19. Mayama C, Araie M, Suzuki Y, et al. Statistical evaluation of the diagnostic accuracy of methods used to determine the progression of visual field defects in glaucoma. Ophthalmology 2004;111:2117–25. 20. Vesti E, Johnson CA, Chauhan BC. Comparison of different methods for detecting glaucomatous visual field progression. Invest Ophthalmol Vis Sci 2003;44:3873–9.

1979

Ophthalmology Volume 114, Number 11, November 2007 21. Tan JC, Hitchings RA. Optimizing and validating an approach for identifying glaucomatous change in optic nerve topography. Invest Ophthalmol Vis Sci 2004;45:1396 – 403. 22. Operation Manual for the Heidelberg Retina Tomograph 3. Version 1.7.0. Heidelberg, Germany: Heidelberg Engineering; 2005. 23. Dreher AW, Tso PC, Weinreb RN. Reproducibility of topographic measurements of the normal and glaucomatous optic nerve head with the laser tomographic scanner. Am J Ophthalmol 1991;111:221–9. 24. Katz J. Scoring systems for measuring progression of visual field loss in clinical trials of glaucoma treatment. Ophthalmology 1999;106:391–5. 25. Nouri-Mahdavi K, Caprioli J, Coleman AL, et al. Pointwise linear regression for evaluation of visual field outcomes and

1980

26. 27. 28.

29.

comparison with the Advanced Glaucoma Intervention Study methods. Arch Ophthalmol 2005;123:193–9. Katz J, Gilbert D, Quigley HA, Sommer A. Estimating progression of visual field loss in glaucoma. Ophthalmology 1997;104:1017–25. Johnson CA, Cioffi GA, Liebmann JR, et al. The relationship between structural and functional alterations in glaucoma: a review. Semin Ophthalmol 2000;15:221–33. Owen VM, Strouthidis NG, Garway-Heath DF, Crabb DP. Measurement variability in Heidelberg Retina Tomograph imaging of neuroretinal rim area. Invest Ophthalmol Vis Sci 2006;47:5322–30. Artes PH, Chauhan BC. Longitudinal changes in the visual field and optic disc in glaucoma. Prog Retin Eye Res 2005; 24:333–54.