In patients with glaucoma, detectable structural loss of the retinal nerve fiber layer (RNFL) has been found to precede apparent visual field dysfunction.1-3 Recent studies have shown that various scanning laser devices, such as the GDx-Nerve Fiber Analyzer (NFA), provide reproducible and objective evaluation of RNFL, and these devices have been used to evaluate differences in retardation between normal and glaucomatous eyes and between normal subjects and patients with ocular hypertension (OHTN).4-9 In addition, polarimetric measurements of the RNFL thickness may detect "preperimetric" glaucomatous optic nerve damage.10

Since the description of "frequency doubling illusion" in 1996, threshold testing has been shown to detect glaucomatous damage earlier than white-on-white conventional perimetry.11-13 FDT screening mode has also shown promise for the detection of glaucoma. This latter technique has a number of additional advantages as a diagnostic device, including rapid assessment, relatively low cost, and easy operation by trained nonphysicians. Using a Humphrey C 24-2 full-threshold algorithm as the definition of glaucoma, the screening mode of FDT has been reported to have a sensitivity of 91% and a specificity of 94% with an abnormal GHT,14 although it is not yet known if this method can detect very early forms of glaucoma including "preperimetric" glaucoma.

Although an association between functional properties and structural alterations of the nerve fiber layer in glaucoma has been described,15-21 attempts to correlate structural changes, as measured by SLP, with functional alterations, as measured by FDT, have been limited in eyes with early glaucomatous damage. We therefore performed a cross-sectional study to evaluate the potential of the FDT screening mode in the early detection of glaucoma by examining RNFL parameters obtained with GDx-NFA in normal controls, glaucoma suspects, and glaucomatous eyes. We also determined the correlation between RNFL retardation, as measured by scanning laser polarimetry(SLP), and FDT findings in our series of glaucoma suspects and glaucomatous eyes.

All study patients and control subjects were consecutively enrolled and examined from June 2002 to May 2003 at the Asan Medical Center, College of Medicine, University of Ulsan, Seoul, Korea. All subjects had best corrected visual acuity of 20/40 or better. If both eyes of a patient met the inclusion criteria, we randomly selected for 1 eye for inclusion in the study. The study followed the tenets of the Helsinki declaration for research involving human subjects, and informed consent was obtained from all participants. We excluded patients with a history of laser surgery, intraocular surgery in either eye, intracranial abnormalities, established diagnosis of neurological disease, or a lesion upon neurological examination. To reduce the influence of age and high prevalence of glaucoma over the age of 40 years, study subjects were restricted to those between 50 and 75 years.

Fifty-three eyes of 53 patients were enrolled as glaucoma suspects, defined as having one or more abnormal locations on FDT C20-1 screening program. They also had suspicious changes of optic nerve head, including cup/disc ratio greater than 0.6, cup/disc ratios higher vertically than horizontally, an asymmetry in cup/disc ratio greater than 0.2 between eyes, or optic disc hemorrhage. The appearance of the RNFL was not taken into account. Each glaucoma suspect had normal open angles on gonioscopy without any evidence of pigmentation, pseudoexfoliation, or other secondary causes of glaucoma on anterior segment examination. Each had normal findings with a Humphrey field analyzer (HFA; Humphrey Zeiss, Dublin, CA, USA) based on our defined criteria. None of the glaucoma suspects had typical visual field defects of neurologic origin including quadrantanopia, hemianopia, altitudinal defects, and central scotomas based on the HFA. Eyes with optic disc abnormalities suggestive of nonglaucomatous origins including congenital anomalies of the optic disk, various types of optic atrophy, and inflammatory and ischemic disorders of the optic nerve head, were excluded from the study. Of the 53 eyes in the glaucoma suspect group, 12 (23%) had IOP greater than or equal to 22mmHg on Goldman applanation tonometry (GAT) at multiple times during the evaluations.

The glaucoma group consisted of 53 eyes of 53 patients with open-angle glaucoma with or without IOP elevation over 21 mm Hg. None had secondary open-angle glaucoma due to pigment dispersion syndrome, pseudoexfoliation, traumatic anterior chamber angle recession, or other secondary causes. Patients with glaucoma had to also fulfill the following criteria: presence of glaucomatous optic neuropathy and associated glaucomatous field based on HFA. Glaucomatous optic neuropathy was defined as either cup/disk asymmetry between fellow eyes of greater than 0.2, rim thinning, notching, excavation, or RNFL defect. Of the 53 eyes in the glaucoma group, 13 (24.5%) had normal-tension glaucoma (NTG) based on the following criteria: an untreated intraocular pressure less than or equal to 21 mmHg at all times over a 24-hour period, measured every 2 hours on admission, a normal open angle of the anterior chamber, the presence of optic nerve as defined above and glaucomatous visual field changes, and no ocular or systemic disorders that could be responsible for optic nerve damage or visual field changes.

Thirty-six healthy volunteers were included in this study as a control group. They were hospital administrative staff members and their relatives. The normal appearance of the optic nerve head in both eyes was confirmed by ophthalmoscopy and stereoscopic photography, and no other significant ocular diseases except for mild age-related cataracts were noted by slit lamp biomicroscopy. None of the control group had a history of IOP elevation greater than 21 mm Hg or ocular trauma. The control subjects also did not have a glaucomatous visual field as defined below. All subjects underwent an HFA C30-2 or 24-2, full-threshold, Glaucoma Hemifield test (GHT) program. All subjects, including controls, had undergone prior visual field examinations more than once, and we used the last visual field test for analysis.

All subjects had to meet a less than 25% rate of false-positive, false-negative responses as well as a 25% fixation loss. None of the glaucoma suspects or normal controls exhibited any specific glaucomatous visual field defect, defined as: (1) three adjacent locations decreased by a sensitivity of 5 dB each or greater with one decreased by at least 10, (2) two adjacent locations decreased by at least 10 dB each, (3) a 10 dB difference across the nasal horizontal meridian at two adjacent locations, or (4) GHT "outside normal limits".

For the FDT testing, all subjects, including normal controls, were tested with distance or bifocal corrective lenses in place. All patients had a refractive error of less than +/- 7 diopters in the sphere or less than +3 diopters in the cylinder. One technician (HWH), trained for the instrument, tested all the subjects, who were given a full explanation together with a chart showing pictures of stimuli. Before the recorded test, the subjects practiced several times. The C-20-1 screening program was used according to the fixed protocol. Test results with good reliability, defined as fixation losses, and false positives of 33% or less were entered for analysis in all subjects. A plot of 17 visual field locations was produced for each eye. If any defect was noted in 17 locations per eye in the first trial after a few practice sessions, the test was repeated 30 minutes later to determine whether there was a learning effect. In these cases, the results of the second test were used as the final FDT data. Three levels of shading indicated mild, moderate, and severe relative loss based on the expected contrast values from a database at multiple levels that would be detected in normal persons of comparable age (99%, 99.5%, and worse than 99.5% level).

For quantification of the defect score, we used the FDT results according to the criteria devised by Quigley, in which the total defect score was assessed by the number of abnormal points in 17 zones, including the central fixation zone times the severity of the measured defect on a scale from 1 (mild) to 2 (moderate) to 3 (severe).14 Regional defect score was calculated according to the same method in each hemifield of eight zones, excluding the central fixation zone. Each patient in the glaucoma suspect group had to have one or more abnormal points somewhere in the field, including the central point, depressed to P value less than 1% with a minimal defect score of 1 and at least one abnormal block.

All subjects were examined with a SLP device(GDx-NFA, Laser Diagnostic Technologies Inc, San Diego, CA, USA) by the same experienced operator (HWH). For each eye, 65,536 retinal locations were measured to create a retardation map corresponding to RNFL measurements over a 15 ×15 retinal area in 0.7 second. While undergoing scanning, the pupils of the subjects were undilated. We used the mean of three good images for analysis, with an average standard deviation of less than 8 um. As defined by the manufacturer, a good image had even illumination and sharp and well-defined edges of blood vessels, as well as lacking red saturation. The optic disc margin was approximated by a circle or ellipse placed around the inner margin of the peripapillary scleral ring by the same operator. The instrument was then placed a measuring circle or ellipse (10 pixels width) 1.75 disc diameters away and concentric with the margin of the optic disc. Defaulted quadrant positions (supplied by the manufacturer) were applied. The peripapillary band was divided into superior and inferior segments of 120 each, a temporal segment of 50, and a nasal segment of 70. The GDx Analysis program automatically calculated for 16 RNFL parameters - symmetry, superior ratio, inferior ratio, superior/nasal, maximum modulation, ellipse modulation, the number, average thickness, ellipse average, superior average, inferior average, superior integral, and deviation from normal- for printout in each of the four quadrants.

All participants underwent FDT testing, HFA examination, and GDx-NFA examination within 1 month of initial clinical evaluation. We compared the FDT defect score, number of abnormal blocks in the total region, FDT test duration, and GDx parameters among the 3 different groups. Since both the defect score and the number of abnormal blocks in the FDT screening testing represented deviations from an age-matched normal database, we correlated each with the average of superior and inferior deviations from normal values (GDx "mean deviation") in the corresponding GDx-NFA printout. Similar correlation analysis was performed between the FDT hemifield defect score and the corresponding superior or inferior quadrant deviation from normal values in the GDx printout. Finally, we performed correlation analysis of the total defect score and the number of abnormal blocks and GDx-NFA indices.

Data were statistically analyzed using software package SPSS (Statistical Package for Social Sciences Inc., version 10.0, Chicago, IL, USA). The Kruskall-Wallis test was used to evaluate differences in FDT and GDx-NFA measurement values among the 3 subject groups. Sensitivity and specificity were obtained based on the FDT abnormality criteria (one or more abnormal locations out of 17 locations). Significance of pairwise comparisons was determined using the Mann-Whitney test. Bonferroni's correction was applied to compare multiple parameters when appropriate. Correlation between FDT data and different parameters of GDx-NFA was assessed with the Pearson correlation coefficient. The multiple regression model was used to assess the GDx-NFA individual parameter with dependent variables. Statistical significance was set at P < 0.05. All numbers are reported as mean +/- standard deviation.

The number, age, refractive error, last IOP, and C/D ratio of all subjects are presented in Table 1. There were no statistically significant differences in mean age, refractive error, or IOP among the three groups. The total mean deviation (MD), pattern standard deviation (PSD), and corrected pattern standard deviation (CPSD) of the 30-2 or 24-2 HFA program, however, were significantly higher in the glaucomatous eyes than in either of the other two groups (P < 0.001 for each; Table 1).

We found significant differences in mean total defect score and number of abnormal blocks on FDT testing among the 3 groups (Table 2). One or more abnormal locations had a sensitivity of 84.9% (45/53) and a specificity of 100% (36/36). In addition, statistically significant differences in the mean FDT testing time were found among the 3 groups (Table 2).

We also observed significant differences in GDx "mean deviation" values from the age-matched normal database, as well as 12 RNFL thickness measurements by GDx-NFA, between the normal controls and each of the other two groups in all pairwise comparisons except "symmetry" (P < 0.001) (Tables 3 and 4). In contrast, no differences in RNFL parameters were found between glaucoma suspects (normal SAP and abnormal FDT) and glaucomatous eyes (abnormal SAP) except "inferior ratio".

Except for "symmetry", all measured values for the glaucoma suspects were intermediate between the measurements for the normal controls and those for glaucomatous eyes. Although the means of the measured variables differed significantly between the normal controls and each of the other two groups, there were noticeable overlaps between controls and glaucoma suspects, and between glaucoma suspects and glaucomatous eyes, for all GDx parameters (Table 3).

In the glaucomatous eyes, we observed negative linear correlations between total FDT defect score and GDx "mean deviation", and between total number of abnormal blocks and GDx "mean deviation" (P < 0.05, Fig. 1). Similar to the full field, FDT defect score in the superior hemifield was negatively correlated with inferior GDx deviation value in glaucomatous eyes (P = 0.01, Fig. 2). In contrast, we found no significant association in the glaucoma suspects between GDx "mean deviation" and either FDT total defect score or the number of abnormal blocks (Fig. 1). Furthermore, we observed poor regional correlation between FDT hemifield abnormal score and corresponding GDx deviation value in these eyes (Fig. 2).

When we analyzed FDT abnormal data and the 12 GDx-NFA parameters (Tables 5, 6), we found that total FDT defect score increased with decreasing "inferior average" in the glaucomatous eyes (r = -0.299, P < 0.05). In addition, the total number of abnormal blocks in FDT increased with increasing GDx "Number" (r = 0.296, P < 0.05), and with decreasing "inferior average", "ellipse average", and "superior average" (P < 0.05 for each), in the glaucomatous eyes. In the eyes of glaucoma suspects, however, we observed no significant correlation between FDT and SLP data. Since there was some correlation between number of abnormal blocks in FDT and various GDx-NFA parameters for the glaucomatous eyes, we used a multiple linear regression model to determine the independent contributions of these parameters. Of the GDx parameters, we found that the most important predictor of FDT total number of abnormal blocks was "inferior average" in the glaucomatous eyes (b = -0.272, SE = 0.104, t = -2.621; P < 0.01).

In glaucoma, it is generally accepted that a significant loss of retinal ganglion cells occurs before functional visual loss appears on conventional achromatic visual field testing. This has led to a search for more sensitive ways of detecting retinal ganglion cell damage at a relatively early stage. Psychophysical examination, such as FDT testing, has been reported to be more sensitive to early visual field loss in eyes with glaucomatous optic neuropathy than standard automated perimetry (SAP). This is accomplished by evaluating the functional properties of a subset of larger retinal ganglion cells (My ganglion cells) which may be more susceptible to glaucomatous optic neuropathy than other ganglion cells.13,22-28 When tested on a group of patients with glaucoma and OHTN, FDT was shown to be superior to SAP for identifying eyes with glaucomatous optic neuropathy. In addition, FDT identified a larger percentage of the abnormal results in the eyes of patients with OHTN than SAP or other psychophysical tests including short-wavelength automated perimetry (SWAP) and motion automated perimetry.22

In patients with open angle glaucoma and visual field defects in only one hemifield, FDT with threshold N-30 program detected visual field changes in approximately 40% of the hemifields that were intact on HFA testing. The Heidelberg Retinal Tomograph (HRT) was shown to detect a significantly smaller rim volume in the corresponding region of the optic disc in eyes with hemifield defects than in control eyes with normal FDT results indicating that FDT may be able to predict abnormal structures in the optic disc or RNFL at an early stage of glaucoma, before a specific conventional visual field defect occurs.13

A suprathreshold screening protocol using FDT has shown reasonably good sensitivity and specificity for the detection of glaucomatous visual field loss.14,23,29,30 In one study, 91% of eyes with abnormal Humphrey perimetric results had two or more abnormal locations in 17 sectors in the FDT C-20-1 screening program,14 while a second study showed that one or more missed locations on the FDT C-20-1 screening program had a sensitivity of 83.5% and specificity of 100% in glaucomatous eyes.23 The program's testing is relatively quick, and test-retest reliability has been shown to be good. To our knowledge, however, the usefulness of the FDT screening protocol had not been evaluated in the early detection of glaucoma, i.e., before visual field defects are present in conventional achromatic perimetry. We therefore tested the ability of this program to predict early glaucomatous structural damage in glaucoma suspects, and the possible functional correlates of structural change in RNFL by using GDx-NFA with FDT screening algorithms (defect score and number of abnormal blocks).

Our glaucoma suspects were enrolled on the basis of an abnormality of reliable FDT screening result and normal achromatic visual field test, while the examiner (MSK) was masked with respect to the status of the optic nerve or RNFL. After enrollment, those eyes with any suspicious nonglaucomatous optic neuropathy were excluded from our study. We hypothesized that these eyes may have sustained prefield glaucomatous damage. When these eyes were evaluated by another diagnostic modality that delineated structural changes in the RNFL, we found that the mean values of all polarimetric parameters, except "symmetry", were intermediate between those of normal controls and glaucomatous eyes.

While the means of measured GDx parameters differed significantly between the glaucoma suspects and control group, there was a marked overlap between glaucoma suspects and glaucomatous eyes. These findings are consistent with an earlier result showing statistical differences in all GDx-NFA values, except for the "symmetry variable", between a "preperimetric" glaucoma group and a control group.10 Our findings also suggest that the use of the FDT screening algorithm may enhance the detection of prefield glaucomatous damage in glaucoma suspects by unmasking morphological changes of the optic disc or RNFL damage before abnormalities appear on achromatic visual field examinations.

Despite its high specificity, the FDT screening protocol using the abnormal criteria of two or more abnormal blocks might be insensitive to early "preperimetric" glaucomatous damage in suspects.14 One study found that only 9% of glaucoma suspects had abnormal FDT with one or more defective blocks, suggesting that the sensitivity and specificity of the FDT screening protocol in detecting early structural damage in "preperimetric" glaucoma suspects require further confirmation.23 On the other hand, detection of an abnormality using the FDT screening protocol, together with suspicious signs of glaucoma in the optic disc despite normal conventional perimetry, suggests that the high specificity associated with the FDT C-20-1 program may strengthen its ability to unmask patients at early stages of glaucoma.

We also investigated whether the FDT screening algorithm could be correlated with RNFL retardation parameters in glaucoma suspects and glaucomatous eyes. Comparisons of achromatic conventional perimetry with SLP have demonstrated correlations between quantitative measurements of the RNFL and visual field loss, as measured both globally and regionally in glaucoma patients.17,31 In our series of glaucoma suspects, however, neither the defect score nor the total number of abnormal blocks could be strongly correlated with the average of inferior and superior GDx deviations from normal values ("mean deviation"). We also found poor correlations between the FDT abnormality score in each hemifield and GDx-NFA measures in the corresponding quadrant. These findings may result from each diagnostic mode targeting on different visual properties of glaucoma. Alternatively, damage to the magnocellular optic nerve fibers, as measured by FDT testing, may not be reflected in the early structural loss as detected by polarimetry, although some histologic evidence suggests that such damage to the magnocellular optic nerve fibers occurs first in the course of glaucoma.32 As observed in glaucomatous eyes, more advanced disease leads to more profound structural loss of RNFL, such that the visual field defect, as detected by FDT, may be better correlated with the structural parameters.

Lack of correlation at an early stage of glaucoma may also reflect high inter-individual variability of RNFL thickness. The wide distribution of normative data as employed by GDx-NFA results in part from the presence of corneal birefringence artifacts which are poorly neutralized by the fixed compensator. This can limit the correlating power between the visual function obtained using FDT and RNFL measurements by GDx-NFA among our glaucoma suspects. This device requires further refinements regarding the influence of anterior segment polarization, individual disc size, split nerve fiber bundle effect, and myopia. Due to the lack of an accepted standard, the diagnostic value of the FDT screening algorithms we utilized may also require further confirmation, as well as a novel classification method based on their efficacy.

While previous studies showed that the highest correlation with Humphrey mean deviation among GDx parameters in perimetric glaucomatous eyes was "The Number",10,31,33 our findings showed that "inferior average" had the highest correlation with the number of abnormal FDT screening blocks, followed by "The Number", "ellipse average", and "superior average". Since "The Number" devised by the neural network is derived from computations based on all available RNFL data, both in normal subjects and glaucomatous eyes with abnormal conventional visual field defects, it may not correlate as well as "inferior average" in the eyes of subjects with early glaucomatous damage. Since our group of glaucomatous eyes included a significant number of eyes with NTG (13/53, or 25%), our finding that "The Number" did not have the highest correlation with the number of abnormal FDT screening blocks may be due to variations in glaucoma severity in our subjects. Other possible explanations include the pattern of glaucoma damage, age, race, study design, and use of different functional tests (i.e. FDT vs. HVF).

Despite the statistical significance of our findings, several factors limited our results. A number of patients in this study were referred for possible glaucoma by outside ophthalmologists as we are engaged in a university-based practice. Although the appearance of the optic nerve head or RNFL was not part of the inclusion criteria, there might have been artificially increased sensitivity and specificity of glaucomatous RNFL change among our patients with defective FDT outcome.

In addition, in this study we used FDT as our standard to quantitatively evaluate the functional glaucomatous damage in the group of glaucoma suspect and glaucomatous eyes. This predefined criterion using FDT results may have artificially increased the classification and inclusion selection bias among study groups. It must also be considered that our results are derived from cross-sectional, as opposed to longitudinal, data, suggesting the need for a longitudinal study to confirm whether glaucoma suspects are more prone to develop conventional glaucomatous visual field defects.

Another limitation of this study is the technique we used for measuring the polarimetry of RNFL with respect to the variable axis and magnitude of corneal polarization. The distribution of the variable corneal polarization axis and magnitude among individuals has been reported to lead to suboptimal RNFL images with elevated retardation or rotated pattern, primarily because, to neutralize corneal birefringence, the GDx-NFA uses a fixed compensator with a magnitude of 60nm and a fast axis oriented at 15 degrees nasally downward.34 Therefore, measurement errors inherent to incomplete compensation of corneal birefringence were not accounted for. However, our effort was to compare the retardation values of the studied eyes with those of normal controls and provide the relative degree of retardation measures among the three groups, as they were equally affected by this inherent error. At the time of this study, we only had access to GDx with fixed corneal compensation (GDx-NFA). The validity of the RNFL measurement would have been enhanced if the inter-individual variability of the corneal polarization could be taken into account by using a variable corneal compensator.

In conclusion, our results suggest that a FDT screening program may aid in the detection of prefield glaucomatous damage, and that our schematic scoring system may predict the relative degree of structural damage in glaucomatous eyes. The lack of correlation between the results of the two diagnostic devices in glaucoma suspects necessitates determination of the role of magnocellular ganglion cells (My cells) in the early glaucomatous disease process. Use of a more refined polarimeter in glaucoma suspects may better elucidate the relationship between the structural alterations observed with SLP and the functional losses detected by the FDT screening protocol. Furthermore, longitudinal studies are needed to determine whether the FDT screening protocol would be useful for predicting future glaucomatous visual field defects.