The Stratus OCT
By Kyung Rim Sung, MD, Kelly A. Townsend, BS; Gadi Wollstein, MD; Hiroshi Ishikawa, MD; and Joel S. Schuman, MD
The clinical detection of glaucomatous progression poses a substantial challenge. The slowly progressive nature of the disease can make it difficult to identify and track. Furthermore, age-related physiological changes and nonglaucomatous disease such as cataract or age-related macular degeneration may confound the identification of truly glaucomatous structural and functional loss during a long-term follow-up period. In addition, there is no gold standard for defining glaucomatous progression, either structurally or functionally. The line between progressive and stable glaucoma is frequently ambiguous, making it problematic to discriminate between true pathological change and unavoidable variability in measurements.
The traditional methods for monitoring glaucomatous progression are the subjective and qualitative assessment of the optic nerve head or retinal nerve fiber layer (RNFL) by ophthalmoscopy or fundus photography. These techniques are insensitive to minute, gradually progressive changes, however, and they are prone to large inter-observer variability. Since optical coherence tomography's (OCT) incorporation into clinical glaucoma assessments, many cross-sectional studies have confirmed its ability to discriminate between glaucomatous and healthy eyes.1-3 The objective and quantitative use of OCT measurements of the retina and optic nerve head may allow clinicians to trace the structural changes due to glaucoma more efficiently than previous subjective descriptive assessments (Figure 1).
Interest in OCT's role in detecting glaucomatous progression has been high since the release of the Stratus OCT (Carl Zeiss Meditec, Inc., Dublin, CA). As a result, the manufacturer has begun developing glaucoma progression analysis software for the device. The company anticipates releasing the first commercially available version in the near future, and more sophisticated methods of analysis are in development. This article focuses on the methods of progression detection, trend analysis and event analysis, that are incorporated into the prototypic version of the software.
OVERVIEW ON TREND AND EVENT ANALYSIS
Trend analysis is defined by the linear regression of RNFL thickness measurements from consecutive scans, and event analysis is defined by changes in thickness from the baseline visit. The crucial component of event analysis is defining the cutoff value for progression. The test/retest variability (reproducibility error) in the measurements of thickness is commonly used as a guideline for setting the threshold value, with changes in thickness beyond test/retest variability regarded as likely pathological change. If an inappropriately low reproducibility error were used as a cutoff, the system's sensitivity to progressive glaucoma might be higher, but the rate of false positives could also be increased (higher sensitivity, lower specificity). Higher cutoffs would be more robust for preventing false positives, but they would also reduce the sensitivity of the system's detection of progression (higher specificity, lower sensitivity).
Event analysis is usually less influenced by the outliers caused by variability in measurements than trend analysis. Because the latter is based on linear regression, one outlier among several visits may affect the overall slope and make it difficult to detect statistically significant changes over time. Trend analysis checks the change in slope of each time point, instead of using standardized cutoffs obtained from population-based data. For that reason, trend analysis could be more useful for detecting individualized progression, because patients may have different rates or phases of progression.
Two additional considerations for detecting disease progression using OCT with either trend or event analysis are variability in measurements due to the quality of the scan and the age-related physiological loss of the RNFL. To correct for those confounders, the prototypic OCT progression software uses a compensatory process to account for the scan's quality and the subject's age.
TESTING THE METHODS
Event Analysis
We analyzed our longitudinal data set using two different cutoffs for event analysis. We defined OCT progression as changes in the RNFL's thickness exceeding the 95% and 99% confidence intervals of the intervisit variability for this parameter4 compared with the baseline thickness in two out of three consecutive visits.
Among 90 eyes with glaucoma and suspected glaucoma, 33 progressed, 12 improved, and 45 were stable when we used the OCT progression criterion of a 95% confidence interval. By the 99% criterion, 24 eyes progressed, five improved, and 61 were stable.
As expected, the 95% confidence interval resulted in larger numbers of progressors, whereas the 99% confidence interval was associated with a smaller number of eyes that improved. Because the RNFL should not thicken in glaucoma subjects, using the 99% confidence interval may be a more appropriate cutoff for detecting progression in this particular data set. Our results confirm the importance of cutoff values for defining progression.
Of the 15 eyes that were defined as progressors by both the 95% confidence interval on OCT and the subjective assessment of visual fields by glaucoma experts, OCT detected progression earlier than visual field analysis in 13 eyes. This finding agrees with the general concept that structural change precedes functional loss.5 It also suggests that OCT may detect glaucomatous progression earlier than functional testing with visual fields.
Trend Analysis
Using our longitudinal data set, we divided the eyes into two groups: progressors and "nonprogressors," defined by the subjective assessment of visual fields. The slope of thinning RNFL over time in subjects not progressing on visual field testing was similar to previously reported thinning due to aging. In contrast, the slope for the progressors was four times faster than in the "nonprogressors," and the difference was highly significant.
The agreement between structurally and functionally defined progression was low in our study. Our findings are in agreement with previous longitudinal studies conducted with a prototypic OCT device and with the Heidelberg Retina Tomograph (Heidelberg Engineering GmbH, Heidelberg, Germany).6-7 The results may indicate that structural and functional progression do not temporally coincide. Moreover, a comparison of two different technologies that measure different aspects of structure and function may be adding complexity to the study, because each modality has its own error in reproducibility and measures different aspects of glaucomatous progression.
CONCLUSION
Assessing disease progression is a crucial component in the evaluation of glaucoma patients. Quantitative assessment with OCT improved our ability to detect minute structural changes over time. Several methods are currently being tested to determine the best method for detecting disease progression.
Kyung Rim Sung, MD, is a research fellow from the UPMC Eye Center, Eye and Ear Institute, Ophthalmology and Visual Science Research Center, University of Pittsburgh School of Medicine. She acknowledged no financial interest in the products or companies mentioned herein.
Kelly A. Townsend, BS, is a research specialist from the UPMC Eye Center, Eye and Ear Institute, Ophthalmology and Visual Science Research Center, University of Pittsburgh School of Medicine, and Department of Bioengineering, University of Pittsburgh School of Engineering. She acknowledged no financial interest in the products or companies mentioned herein. Ms. Townsend may be reached at townsendka@upmc.edu.
Gadi Wollstein, MD, is Assistant Professor of Ophthalmology and Director of the Ophthalmic Imaging Laboratory at the UPMC Eye Center, Eye and Ear Institute, Ophthalmology and Visual Science Research Center, University of Pittsburgh School of Medicine, and Assistant Professor of Bioengineering, University of Pittsburgh School of Engineering. He has received research funds from Carl Zeiss Meditec, Inc. Dr. Wollstein may be reached at (412) 647-0325; wollsteing@upmc.edu.
Hiroshi Ishikawa, MD, is Assistant Professor, Departments of Ophthalmology and Bioengineering, University of Pittsburgh Schools of Medicine and Engineering, and he is Director, Ocular Imaging Center, UPMC Eye Center in Pittsburgh. He acknowledged no financial interest in the products or companies mentioned herein.
Joel S. Schuman, MD, is the Eye and Ear Foundation Professor and Chairman of Ophthalmology at the Eye and Ear Institute of the University of Pittsburgh School of Medicine. He is also Director of the UPMC Eye Center and Professor of Bioengineering at the University of Pittsburgh School of Engineering. During the past 3 years, Dr. Schuman has received research funding, research equipment, honoraria, and/or payment of faculty travel expenses from Carl Zeiss Meditec, Inc., and Heidelberg Engineering, Inc. He receives royalties from intellectual property licensed by M.I.T. to Carl Zeiss Meditec, Inc. Dr. Schuman may be reached at (412) 647-2205; schumanjs@upmc.edu.
The HRT3
By Robert J. Noecker, MD, MBA
The use of the Heidelberg Retina Tomograph (HRT; Heidelberg Engineering GmbH, Heidelberg, Germany) as an imaging tool in glaucoma was validated in the ancillary study of the Ocular Hypertension Treatment Study.1 In that study, the HRT was shown to be a useful tool for predicting the development of glaucoma in an ocular hypertensive population. Clinically, the HRT is normally used in two ways. At baseline, the characteristics of the optic nerve in question are compared to a normative database. Clinicians subsequently use the data acquired over time to determine if specific features of the optic nerve are remaining stable or are changing. This article focuses on the newest version of the HRT platform, the HRT3 (Heidelberg Engineering GmbH).
IMPROVEMENTS ON THE HRT
The HRT3 has many features that enable the system to analyze the optic nerve head and retinal nerve fiber layer (RNFL) more accurately than the original platform. One of its most important features, however, is its software's ability to analyze images and data obtained by prior versions of the HRT. The HRT3's backward compatibility makes it an extremely powerful clinical tool for evaluating glaucomatous progression.
Compared with previous versions of the HRT, the HRT3 provides more feedback to the operator about the quality of the acquired image. A technician can thus repeat a test that is poor in quality so as best to assist with clinical decision making.
The HRT3 software also features an expanded normative database that permits the specification of a given patient's ethnicity. Reference data are available for individuals of Caucasian, African, and South Asian (Indian) descent. In the future, Hispanic and East Asian databases will be included (For more on the significance of ethnic-specific data in glaucoma assessments, see the article by Rohit Varma, MD, PhD, on page 40).
THE PRINTOUT
The top portion of the standard HRT3 printout (Figure 1) includes information about the patient. The most important elements are the patient's assigned ethnic database, the assessment of image quality, and the optic disc's size. Because all subsequent analyses are based on these parameters, it is important to confirm that the information is correct.
Image Quality
The image-quality assessment is indicated by a rating from excellent to poor with the corresponding standard deviation noted in parentheses. This presentation is an improvement on prior versions of the HRT, which required clinicians to make interpretations based only on the raw data.
The Optic Nerve
The HRT3 calculates the size of the optic disc by an algorithm and notes the measurement at the top of the printout in square millimeters. Optic nerves are classified as small if their area is less than 1.6 mm², average if it is between 1.6 and 2.5 mm², and large if it is greater than 2.5 mm². The system bases further analyses on comparisons between the subject's and similarly sized optic nerves.
The images of the optic nerves are represented side by side on the printout, and the HRT3 checks the subject's eyes for asymmetry. It also performs a statistical analysis that compares the parameters of the cup against the normative database. Values of P>.05 are indicated as within normal limits, those of P<.05 are labeled borderline, and P<.001 are flagged as outside normal limits.
The HRT3 also compares the parameters of the optic disc's rim against the normative database and between the patient's eyes. Statistical differences are noted in the manner just described. If clinicians so desire, they can have the HRT3 perform Moorefield's Regression Analysis, with both a global assessment and an assessment of sectoral regions with an indication of consistency with the normative database.
The RNFL
The bottom section of the standard printout displays information about the peripapillary RNFL. The profile of the patient's RNFL is a new feature with the HRT3. The profile is plotted against the normative database with the corresponding statistical comparison both against the normative database and each of the subject's eyes.
GLAUCOMA PROBABILITY SCORE
The glaucoma probability score (GPS) is a new evaluative tool with which the HRT3 performs a statistical analysis of the 3-D shape of the optic disc and RNFL. Because the program uses no contour line, it depends less on the operator's skill than did previous versions of the HRT. The GPS software calculates the probability of damage consistent with glaucoma. It classifies the results based on probability values as outside normal limits (probability > 64%), borderline (probability > 28%), and within normal limits (probability < 28%). The GPS report looks similar to the standard printout, and includes information on the patient, a distribution of parameters, a chart of probability values, and a probability bar chart showing the classification of the results.
DETECTING PROGRESSION
The topographic change analysis (TCA) is a powerful tool that detects glaucomatous progression by statistically analyzing change at each pixel location in an image over time (Figure 2). If the detected alteration exceeds baseline variability or noise, the TCA software designates it statistically significant. The analysis for progression or change is not dependent on where a contour line is placed or the presence of a reference plane.
The data are presented on a change probability map and also on a grayscale image of the optic nerve and RNFL. Focal regions where there has been a significant and consistent change are displayed in red and green. Red areas indicate regions of decreased height, and green regions note those with increased height. A color gradient indicates the magnitude of change, with darker shades representing larger magnitudes.
When viewing the TCA results on the screen, clinicians will see the topographic map, the change-probability map, and the height-change map. From a practical standpoint, the height-change map is most useful clinically, because loss of height in the relevant areas of the optic nerve and the peripapillary RNFL can easily be correlated to the clinicians' observations and changes in functional tests.
The TCA software identifies changes in both area and volume. Its software model is validated against a 10-year database of cases in which glaucoma progressed and can be customized to track specific areas of interest.
The progression printout displays the results of the patient's baseline examination (Figure 3A) and then the follow-up examinations in order below it. Red areas on the images indicate significant changes from baseline (Figure 3B). In addition, the report provides a trend-analysis graph that shows the data more quantitatively over time.
The progression software can be manipulated to extensively re-analyze images and data obtained with earlier versions of the HRT3 and integrate them into current data sets. Clinicians can reset a patient's baseline after a significant intervention such as surgery, and use the new reference values to assess changes from that point forward. After the HRT obtains three images from the same patient, its software can automatically detect changes from the preset baseline and provide a reliable quantitative analysis of the changes. The rapid rate at which images are captured reduces the variability introduced by the manual realignment of the device between serial images.
Future versions of the HRT3 will superimpose the structural information they acquire over functional data from visual field tests, thereby making it easier for clinicians to evaluate the relationship between these two aspects of glaucoma.
CONCLUSION
The HRT3 represents a definite improvement over prior versions of the confocal scanning laser. Its ability to acquire reliable images and the inclusion of software that is compatible with images obtained with older versions of the instrument, provide a considerable amount of information that can help clinicians analyze and track glaucomatous changes. Other useful features include improved normative databases that allow clinicians to customize their analysis of each patient, as well as new algorithms for progression analysis expand the methods available to detect change in different types of optic nerves. Finally, the HRT3's user-friendly printouts supply information about both eyes together. These improvements make the HRT3 a powerful tool for practitioners who diagnose and treat glaucoma.
Robert J. Noecker, MD, MBA, is Director of the Glaucoma Service and Associate Professor/Vice Chair of the Department of Ophthalmology at the University of Pittsburgh. He has received honoraria from Heidelberg Engineering, Inc., and has received grant/research support from Carl Zeiss Meditec, Inc. Dr. Noecker may be reached at (412) 647-5753; noeckerrj@upmc.edu.
THE GDX SCANNING LASER POLARIMETER
By Robert D. Fechtner, MD
The GDx (Carl Zeiss Meditec, Inc., Dublin, CA) is a scanning laser polarimeter that takes advantage of the retina's polarizing structure to measure the thickness of the retinal nerve fiber layer (RNFL). Since the first commercial scanning laser polarimeter was introduced in the late 1980s, it has evolved into a valuable tool for helping clinicians evaluate the RNFL and detect glaucoma.
Carl Zeiss Meditec, Inc., has developed two new features for the GDx: the recently introduced Review with Guided Progression Analysis (GPA), and the forthcoming Enhanced Corneal Compensation (ECC) software. This article describes the current state of the GDx's technology, and discusses how these upgrades will be helpful in my practice.
EVALUATING THE RNFL
Physicians tend to overlook the GDx because its only application is to examine the RNFL. I believe it is a highly relevant technology, because it makes it easier to evaluate the RNFL clinically than with a slit lamp or a photograph. The GDx has progressed through several generations of thoughtfully and systematically designed hardware. I like to think of it as technology that can extend our senses and show us structural changes that we might miss otherwise.
The polarizing properties of the cornea, and to a lesser extent, the crystalline lens, affect the accuracy of results obtained by scanning laser polarimetry. The device must compensate for these structures to extract the signal from the RNFL. Early versions of the scanning laser polarimeter applied an average fixed compensation to the RNFL scan that worked in many, but not all eyes. The inclusion of variable corneal compensation in the current version of the GDx has increased the number of eyes that can be scanned adequately, but some still cannot be imaged.
To further reduce the incidence of excessive scanning artifact, Carl Zeiss Meditec has developed ECC, a software-only upgrade that the company expects to make available in 2008. Based on my experiences with the prototypic ECC software,1 I believe that its introduction will further expand the number patients who can be scanned successfully with the GDX.
THE GDX AND PROGRESSION SOFTWARE
Glaucoma specialists have relied on progression software to detect changes in visual fields for some time. In fact, it is only because of these statistical tools that we can have confidence in our ability to identify progressive visual field loss.
A similar strategy has been developed for the GDx to statistically analyze glaucomatous progression in the RNFL. To detect changes in this structure, we require specific information. First, we must estimate the degree of variability between retinal scans. Next, we must be able to determine if the differences observed between scans over time exceed the expected limits of intertest variability and likely reflect an actual structural change. The serial analysis software included in the current version of the GDx is rather fundamental, because it shows areas of change but cannot perform this kind of statistical analysis.
The Review with GPA provides two modes of statistical analysis. The Fast Mode compares the results of single scans with variability data collected from a test population. This allows the GPA to be used with retrospective normative data. In contrast, the Extended Mode compares triple scans. To use this mode, the operator obtains three scans of the patient's retina during one imaging session. The GPA software uses these scans to calculate individual patient variability, a strategy that should improve the GDx's performance. Both modes require a minimum of three imaging sessions to detect progression, and both can analyze scans obtained with variable corneal compensation or ECC, although these cannot be mixed. If single and triple scans exist, they can be analyzed together using the GPA's fast mode. Carl Zeiss Meditec, Inc., is providing a software upgrade for the instrument to facilitate the triple scanning function.
Imaging tests are less demanding on patients than visual field tests and thus can be repeated more than once a day. This aspect of the GDx allows users to confirm possible glaucomatous progress as soon as it is detected. Although more than one baseline measurement could theoretically be obtained during a single imaging session, the optimal interval for obtaining these reference tests is not known. The manufacturer recommends taking baseline readings (using the triple-scanning function) 2 or 3 months apart.
The Review with GPA software is not incorporated into the GDx's hardware, but instead runs on a separate personal computer. This configuration provides flexibility for viewing images in multiple examination lanes and allows clinicians to store data in a central location. Figures 1 to 3 show standard GDx GPA printouts.
IMPACT ON PATIENT CARE
We may grow complacent as we see our patients over time, because currently available diagnostic techniques are not sensitive enough to detect the subtle progressive changes caused by glaucoma. The inclusion of GPA in the Humphrey Field Analyzer (Carl Zeiss Meditec, Inc.), showed that the damaged areas in some of my patients' visual fields were probably progressing, and not fluctuating as I had believed. Before I had access to this software, I did not have a useful tool to look for progression in these areas of damage.
The Review with GPA software for the GDx is the first statistical tool developed to help us detect changes in the RNFL over time. This technology can either reassure us that the RNFL is stable or give us an early warning that we need to institute more aggressive treatment.
LEARNING CURVE
As clinicians, we are continually challenged by the lack of a gold standard against which we can evaluate the accuracy and utility of various progression software programs. The GDx, Stratus OCT (Carl Zeiss Meditec, Inc.), and the Heidelberg Retina Tomograph (HRT; Heidelberg Engineering GmbH, Heidelberg, Germany) all use different strategies to measure structures. Even with the most sophisticated software, these devices can only tell us if structural change is likely to have occurred within a specific statistical definition. It is still up to us to evaluate all of the data obtained by diagnostic devices and determine if the statistical changes are detecting glaucomatous progression, and what level of change is clinically meaningful. The value of the devices ultimately depends on our ability and experience as clinicians to interpret the data they collect, and how we use this information to preserve our patient's vision for as long as possible.
The GDx is a robust tool that helps me better see the RNFL. GPA software will help me look for statistical changes in the RNFL that represent glaucomatous progression, just as the GPA software on the Humphrey Field Analyzer helps me detect progressive visual field loss.
Additional support for this article was provided by Zyg Kunczynski from Carl Zeiss Meditec, Inc.
Robert D. Fechtner, MD, is Professor of Ophthalmology at the Institute of Ophthalmology and Visual Science, New Jersey Medical School, Newark. He is a consultant for Carl Zeiss Meditec, Inc. Dr. Fechtner may be reached at (973) 972-2030; fechtner@umdnj.edu.
