Structure and Function in Patients with Glaucomatous Defects Near Fixation

Abstract

Purpose

To assess relations between perimetric sensitivity and neuroretinal rim area using high-resolution perimetric mapping in patients with glaucomatous defects within 10 degrees of fixation.

Methods

One eye was tested in each of 31 patients with open angle glaucoma enrolled in a prospective study of perimetric defects within 10 degrees of fixation. Norms were derived from 110 control subjects free of eye disease ages 21 – 81. Perimetric sensitivity was measured using the 10-2 test pattern with the SITA Standard algorithm (HFAII-i, Carl Zeiss Meditec), stimulus size III. Area of the temporal neuroretinal rim was measured using the Heidelberg Retinal Tomograph (HRT III). Decibel (dB) values were converted into linear units of contrast sensitivity averaged across locations corresponding to the temporal rim sector. Both measures were expressed as percent of mean normal and the Bland-Altman method was used to assess agreement. Perimetric locations corresponding to the temporal sector were determined for six different optic nerve maps.

Results

Contrast sensitivity was moderately correlated with temporal rim area (r² > 30%, p < 0.005). For all six optic nerve maps, Bland-Altman analysis found good agreement between perimetric sensitivity and rim area with both measures expressed as fraction of mean normal, and confidence limits for agreement that were consistent with normal between-subject variability in control eyes.

Conclusions

Using high-resolution perimetric mapping in patients with scotomas within 10° of fixation, we confirmed findings of linear relations between perimetric sensitivity and area of temporal neuroretinal rim, and showed that the confidence limits for agreement in patients with glaucoma were consistent with normal between-subject variability.

Glaucoma is the second leading cause of blindness worldwide^{1, 2}, and assessment of severity of glaucomatous damage is important in diagnosis and management. Perimetry is used to measure severity of visual loss, and optic nerve examination frequently finds thinning of the neuroretinal rim or retinal nerve fiber layer (RNFL) in patterns corresponding to the visual field loss. However, variability associated with both perimetric and anatomical measures limits the ability to effectively combine these complementary measures of disease severity.^{2, 3}

In the closing decades of the 20^th century it was widely believed that optic nerve and RNFL defects preceded perimetric defects⁴. This was explained by the “functional reserve” doctrine, which asserted that the normal healthy state has an excess number retinal ganglion cells so that visual function will remain unchanged until ganglion cell loss has become profound, and that the earliest stage of glaucoma is “pre-perimetric”. Over the past decade, this doctrine has been challenged by hypothesis-based studies^5–9, and additional studies that have reported that perimetric defects can precede anatomical defects in early glaucoma^10–12. A comparison between results of multi-focal electrophysiology and perimetry lead to the “simple linear model” of relations between ganglion cells and perimetric sensitivity¹³, which opposes the doctrine of functional reserve and posits that on average perimetric losses are equal to ganglion cell losses. This model has been extended to structure-function relations¹⁴, and posits that between-subject variability in people free of visual disorder is the primary cause of discordance between structural and functional measures of disease severity.¹⁵

One of the seminal studies on linear sensitivity⁶ was a comparison by Garway-Heath and colleagues of rim area of the temporal half of the optic disc with visual field sensitivity for the central 16 points of the 24-2 test pattern, including both patients with glaucoma and people free of eye disease. This study found that the appearance of a functional reserve could be explained as a statistical artifact from using log units for perimetric sensitivity and linear units for anatomical measures. When perimetric measures were converted to linear units, the relations became linear. A recent study¹⁶ by Racette and colleagues attempted to replicate this result in 385 people, assessing all six sectors of the optic disc rather than the temporal half of the disc as in the original study. For many sectors the linear prediction had statistical support, but results were ambivalent for the temporal sector versus the central 16 visual field locations. Both of these studies relied on linear regression versus polynomial regression, an approach whose statistical validity is limited by the fact that there is not an independent variable for these clinical measures. We developed a method of quantitative analysis, based on the simple linear model, which avoids this limitation.

The temporal sector is of particular importance in assessing severity of glaucomatous damage, as the area spanned by the central 16 locations includes the macula. A scotoma that enters the macula can seriously affect activities of daily living. Therefore we felt that it was appropriate to replicate the original study⁶, focusing only on the temporal sector, using 68 test locations with the 10-2 test pattern rather than 16 from the 24-2 test pattern. We only recruited patients with visual field defects on the 10-2, and used six different maps relating 10-2 test locations with the temporal sector in order to assess robustness of the analysis of agreement.

METHODS

Subjects

Thirty-one patients with open angle glaucoma were recruited prospectively for this study, as part of a multifaceted study of 10-2 fields at the SUNY Glaucoma Institute. Inclusion criteria were: glaucomatous optic neuropathy with corresponding visual field defects on the 24-2 test pattern, experienced with automated perimetry, clear ocular media, visual acuity of 20/30 or better, pupils > 2mm in diameter, spherical refraction within ±5.0 diopters and cylinder correction within ±3.0 diopters. Exclusion criteria were cataract, concomitant disease or medication known to affect the visual field or macular function, optic disc abnormality (other than glaucomatous). From this patient pool, we recruited only patients who had repeatable perimetric defects within 10 degrees of fixation, and a history of reliable visual fields.

There were 16 females and 15 males. The median spherical equivalent of refractive error was 0.00 diopters, and ranged from −3.75 diopters to +3.75 diopters; only 7 patients had refractive errors greater than ± 2.00 diopters.

Ethnicity and race were recorded as separate categories. Five patients identified themselves as of Hispanic ethnicity, and 26 as of non-Hispanic ethnicity. One patient identified their race as American Indian, two as White, and 28 as African-American. Their mean ± SD age was 66 ± 12 years. For the HRT-III, their mean disc area was 2.44 ± 0.58 mm² and their mean temporal rim area was 0.15 ± 0.08 mm². For the 24-2, their MDs averaged −8.1 ± 6.9 dB and PSDs averaged 7.7 ± 5.4 dB; for the 10-2 their MDs averaged −10.6 ± 8.1 dB, and PSDs averaged 6.5 ± 3.6 dB. For these 10-2 fields, no patient had a false positive rate greater than 15%, the median was 1%, and only 4 patients had a value higher than 5%. No patient had a fixation loss rate greater than 27%, the median was 0%, and only 7 patients had a rate higher than 7%. No patient had a false negative rate greater than 14%, the median was 0%, and only 3 patients had a rate higher than 5%.

The study was conducted following the tenets of the Declaration of Helsinki. Written informed consent was obtained from each participant prior to testing, after explaining the procedure and Goals of the experiment. The protocol was approved by the Institutional Review Board of SUNY State College of Optometry.

Apparatus

Confocal scanning laser ophthalmoscopy was performed using the Heidelberg Retina Tomograph 3 (HRT III). The contour line for the HRT was drawn by one of two authors (AS and MWD), who outlined the boundaries of the optic nerve using the three-dimensional view from the HRT III. We used the reported area of the neuroretinal rim for the temporal sector of the nerve, spanning −45° to 45°.

Macular perimetry was performed using stimulus size III white-on-white (HFA II-i, Carl Zeiss Meditec), with the 10-2 test pattern and the SITA Standard algorithm. In contrast to the 6-degree separation between test points used in the 30-2 and 24-2 test patterns, the 10-2 test points are separated by 2 degrees. We used a size III stimulus (0.43° diameter), flashed on for 200 msec, as used in prior studies^{6, 16} with the 24-2 test pattern.

Maps of Visual Field to Optic Disc

We used six different maps to determine locations of the 10-2 test pattern that correspond to the temporal rim sector. For each patient and map, we computed the arithmetic mean of the perimetric sensitivities for these locations, which varied in number from 34 to 68 across maps. This arithmetic mean is the average of sensitivities expressed as linear contrast sensitivity (a unitless measure, expressed as the reciprocal of Weber contrast)^9,17,18, and we refer to this as the contrast sensitivity corresponding to the temporal sector of the optic disc.

The models of Garway-Heath and colleagues⁶ and Hood and Kardon¹⁹ for structure-function relations both use arithmetic means and rely on the map produced by Garway-Heath and colleagues²⁰, which we used as our primary map. The Garway-Heath et al. model gives more weight to the macular locations, while the Hood-Kardon model weights all locations equally as defect relative to mean normal. As there is much less effect of eccentricity on perimetric sensitivity for the 10-2 test pattern than for the 24-2 test pattern, we did not to weight locations based on eccentricity or relative to mean normal.

The original publication of the Garway-Heath et al. map²⁰ was for an older version of the HRT software, where the temporal sector spans from −49° to + 40°. We used the updated version of the map in the second publication⁶, which corresponds to current HRT software and the temporal sector spans from −45° to +45°. This change removed a superior nasal location from the temporal sector of the optic disc, leaving only 5 locations from the 24-2 that mapped to the temporal sector. Therefore Garway-Heath et al.⁶ averaged across the 16 locations within 13 degrees of fixation, and compared this mean to the rim area for the entire temporal half of the optic disc.

The 10-2 test pattern includes 68 field locations, all within the original region of the visual field sampled by the 16 locations used by Garway-Heath et al.⁶. By removing selected visual field locations, we were able to halve the region of the optic disc used, comparing the means to the ± 45° temporal sector rather than using the entire temporal half of the disc. The six maps removed from 0 to 34 locations for the arithmetic mean, as illustrated in Figure 1 and described below. We derived these maps from published maps for the 24-2.

Figure 1.

Locations of the 10-2 perimetric test that map to the temporal optic disc sector, in right eye format so that nasal visual field is represented by negative values on the x-axis. Black circles indicate locations excluded from the temporal sector for this map. For the Ferreras map (middle right) grays are used as well, representing partial weightings of 0.25 (dark gray) and 0.5 (light gray), compared to 1.0 for white circles and 0.0 for black circles. Left column (top to bottom) shows Full 68, Gardiner, and Jansonius maps. Right column (top to bottom) shows Harwerth, Ferraras, and Garway-Heath maps.

Full 68

This simple model assumes that most of the locations correspond to the temporal sector, and simply sums across all 68 locations rather than relying on a map. The model assumes that for an individual patient all but a fraction of these 68 locations will correspond to locations within ±45°, and that normal between-subject variability will cause that fraction to vary from person to person. For convenience we refer to the remaining maps by the name of the first author of the study that produced them.

Harwerth

This 10-sector map by Harwerth and colleagues²¹ has sectors 1 & 10 spanning ±36°, and projecting only to the temporal visual field, so the 34 nasal visual field locations are removed.

Gardiner

This 24-2 map by Gardiner and colleagues²² has minimal contribution to the temporal sector from superior-nasal visual field, and similar levels of contribution from the other three sectors, Therefore only the 17 superior-nasal locations are removed.

Ferreras

This map by Ferreras and colleagues²³ assigns, to each of 52 visual field locations of the 24-2, probabilistic weightings for each of 12 sectors of peripapillary retinal nerve fiber layer spanning 30° each. The central 3 sectors at the temporal side of the optic disc span the same 90° that the HRT-III “temporal rim” spans. The central sector spans ± 15°, and has no clear weighting for any visual field location. The sector from −45° to −15° has significant weighting for the macular visual field, while the sector from +15° to +45° has weak weighting for the macular locations. For this map, the temporal sector is only assigned full weight for the 17 superior-temporal locations of the 10-2, half weight to 13 nasal locations, and quarter weight to 9 inferior-temporal locations. The other 39 locations are removed.

Jansonius

This map by Jansonius and colleagues²⁴ was derived by tracing RNFL bundles to a mean angle on the optic nerve in normal eyes, assigning angles of −135° to ±135° for the sector we are examining. This map excludes 22 locations: 13 superior nasal locations, 4 superior temporal locations, 3 inferior nasal locations and 2 inferior temporal locations.

Garway-Heath

We constructed this using the retinotopic maps of Wyatt²⁵ and Garway-Heath et al.⁶, in collaboration with Harry Wyatt. This map uses a subset of the locations in the Jansonius map, discarding sensitivities for 33 locations: all 17 locations in the superior-nasal quadrant of the 10-2 test pattern, 8 locations in the inferior-nasal quadrant, and 4 each in the superior and inferior temporal quadrants.

Statistical Design

Our goal was to compare quantitative measurements of the neuroretinal rim area from the temporal sector of the HRT III and corresponding 10-2 sensitivities. Each of the six maps in Figure 1 was used to compute the arithmetic mean for each patient, using contrast sensitivities that were calculated from Weber contrasts⁹ for the threshold stimuli at each location that the map assigns to the temporal sector. Our primary analysis was with the Garway-Heath map, and we repeated all analyses with each of the other five maps.

For direct comparison with the study by Garway-Heath et al.⁶ we computed the correlation between rim area and contrast sensitivity and evaluated r², which was their outcome measure. The value of r² is a property of the data and does not depend on which variable is on the x-axis, but linear regression cannot be properly conducted without an independent variable. For a more comprehensive analysis of agreement, we used the Bland-Altman method²⁶, as we have in previous studies^{9, 17, 18}. As compared to linear regression, the Bland-Altman method does not require an independent variable for validity and is not affected by choice of x-axis, or by which measurement has lower variability. The difference between the two measures was plotted versus the mean of the two measures, and correlation was used to determine whether the difference score was dependent on the mean level of damage. If the correlation did not find a dependence (p > 0.05), then the mean of the difference score was computed as the mean difference between the measures and a two-tailed t-test was used to determine whether the mean difference was non-zero (p > 0.05). The standard deviation (SD) of the difference scores (or of the residuals for linear regression if the correlation was significant) was used to compute 95% limits of agreement for the Bland-Altman analysis, computed as ± 1.96 SD.

In order to provide common units for both measures, as needed for Bland-Altman analysis, we converted both measures to percent of mean normal¹⁷. Norms for the 10-2 were obtained from the SUNY Glaucoma Institute normative database (55 control subjects, ages 21–81), preliminary data published in Wyatt et al.²⁷; no control subject had a false positive rate greater than 15% (median 0%), false negative rate greater than 14% (median 0%), or fixation loss rate greater than 27% (median 0%). The mean PSD was 1.27 ± 0.18 dB for this group of controls, and the MD was −0.9 ± 0.8 dB. Linear contrast sensitivity for the temporal sector showed a significant effect of age (r = −0.55, p < 0.0001), with variability of the residuals independent of age. Therefore the linear regression line was used to compute mean normal as a function of age. This was very similar to machine norms for effect of age, except that our norms were systematically lower than machine norms. Had we used machine norms, perimetric sensitivities would have all been slightly smaller fractions of mean normal.

Norms for temporal rim area were obtained from an ongoing longitudinal study (SUNY Glaucoma Institute in collaboration with Indiana University School of Optometry) consisting of 73 control subjects ages 23–85 with global disc areas of 0.9 to 4.8 mm², all with SD < 30 microns on the HRT. Multiple regression yielded little effect of disc area and age, with r² < 6%, p>0.20. Therefore we did not factor into account the size of the disc or the patient’s age. The global rim area for this group was 1.66 ± 0.48 mm², slightly larger than machine norms; if we had used machine norms then rim values for our patients would most likely have been a larger fraction of mean normal.

In a secondary analysis we fit the data with the Hood-Kardon model¹⁹ adapted for the Bland-Altman analysis. This model assumes that there is a non-neural component that is unaffected by ganglion cell loss, and that contrast sensitivity on average declines linearly with the neural component. This model has a single parameter, the magnitude of non-neural component, and the Bland-Altman plot was fit with the model to estimate the non-neural component; Z-scores were generated from the mean and standard error (SE) for the parameter fit, and the one-tailed 95% confidence limit was computed as the mean plus 1.64 times the SE.

RESULTS

Figure 2 shows a scatterplot of rim area versus contrast sensitivity for our data and Garway-Heath et al.’s data)⁶, where contrast sensitivities for both studies were computed with the Garway-Heath map. Our control group (circles) had rim area about half that of Garway-Heath et al.’s control group (squares), as expected because we assessed rim area for half as much of the optic disc. Mean normal contrast sensitivity for our control group was almost twice that for Garway-Heath et al., consistent with their use of all locations in a 16° by 16° square centered on fixation versus our use of a subset of locations within 10° from fixation. Correlation yielded r² values of 4% (p > 0.23) for the Garway-Heath et al. data and 31% (p < 0.005) for our data. With the other five 10-2 maps, the r² values for our data were slightly larger: 41% for Full 68, 34% for Harwerth, 38% for Gardiner, 46% for Ferraras, 38% for Jansonius. Note that the lines in Figure 4 are not regression lines (as there is not an independent variable), but are simply lines passing through zero and the points for mean normal.

Figure 2.

Perimetric sensitivity versus area of neuroretinal rim for the temporal optic disc sector, for our data (circles) and the data of Garway-Heath et al. (squares). Large solid symbols show mean values for the norms used for each group, and error bars show two-tailed 95% confidence limits for normal. Diagonal lines show predictions for equal loss, starting at mean normal and ending at zero. The size of each symbol from our study corresponds to the disc area for each subject (larger circles for larger disc areas and smaller circles for smaller disc areas). The shade of each symbol from our study represents the spherical equivalent of refractive error (darker shades for more negative refractions, lighter shades for more positive refractions). Perimetric sensitivity is expressed as contrast sensitivity, which is the reciprocal of Weber contrast of the threshold stimulus and is a unitless measure.

Figure 4.

Perimetric sensitivity versus area of neuroretinal rim for the temporal optic disc sector, expressed as percent of mean normal , along with Bland-Altman regression lines and confidence limits. Perimetric sensitivity was computed for each of the six maps in Figure 1: left column (top to bottom) shows Full 68, Gardiner, and Jansonius map; right column (top to bottom) shows Harwerth, Ferraras, and Garway-Heath maps. Symbol size increases with global disc area, and the shade of each symbol represents the spherical equivalent of refractive error. There was no apparent influence of disc size or refractive error on the results of Bland-Altman analysis for any of the maps.

For both our data and the Garway-Heath et al. data, the Bland-Altman analysis is shown in Fig. 3. Mean differences were not different from zero (t < 0.7, p > 0.54), with mean ± SD of −0.03 ± 0.28 for the data from the Garway-Heath et al. study and +0.01 ± 0.25 for our data. For both datasets there was no dependence of difference score on the mean level of damage (r² < 3%, p > 0.37). Similar results were obtained for the five alternate maps.

Figure 3.

Bland-Altman analysis, computed from the data in Figure 2 converted to percent of mean normal: our data are shown as circles and the data of Garway-Heath et al. are shown as squares. The solid lines show regression lines and 95% confidence limits for our data, and the dashed lines show regression lines and 95% confidence limits for the Garway-Heath et al. data. For each datatset, the slope and intercept of the regression line was not significantly different from zero (t < 0.7, p > 0.54). For our data, symbol size increases with global disc area, and the shade of each symbol represents the spherical equivalent of refractive error. There was no apparent influence of disc size or refractive error on the analysis of agreement.

For completeness the figure shows the regression lines and their confidence limits, but as the means and slopes were not significantly different than zero, we conclude that the mean difference was not different than zero and the 95% confidence limits were approximately ± 0.5. Therefore the two measures are on average in agreement, but for any given patient there can be considerable difference between the two measures. For instance, a patient with a mean of 0.5 and a difference score of 0.5 could have rim area 0.75 times mean normal (within normal limits) and contrast sensitivity at 0.25 times mean normal (well below normal limits), or vice versa.

Figure 4 shows scatterplots of perimetric sensitivity versus rim area for all six maps, with both rim area and perimetric sensitivity expressed as fraction of mean normal, and the diagonal lines showing the mean and confidence limits from the Bland-Altman analysis. The 95% confidence limits from the Bland-Altman analysis were consistent with the 95% confidence limits for normal variability in control eyes, as shown by the Bland-Altman confidence limits for agreement (dashed lines) intersecting the confidence limits for normal variability (error bars).

In secondary analysis with the Hood-Kardon model, for all six models we obtained Z-scores less than 0.3 for the non-neural component being greater than zero (p > 0.48); the one-tailed 95% upper limit the non-neural component was 0.14 to 0.15 times mean normal across the six maps.

DISCUSSION

We confirmed prior findings that perimetric sensitivity and rim area on average show equal amounts of loss in patients with open-angle glaucoma. The Garway-Heath et al.⁶ and Racette et al.¹⁶ studies included patients with a wide range of disease severity as well as glaucoma suspects and healthy controls free from glaucoma. When the healthy controls and glaucoma suspects were removed from the analysis, the correlation became insignificant for the 40 patients studied by Garway-Heath et al. By comparison, for our study with 31 patients the correlation was much larger and reached significance. The stronger correlation for our data may reflect our focus on patients with visual field damage within 10° of fixation and our use of high-resolution perimetric mapping, which allowed us to restrict our analysis to the temporal quarter of the optic disc.

All of the subjects in the Garway-Heath et al. study, and the vast majority of the subjects in the Racette et al. study, were of racial and ethnic backgrounds that would class them as “White, not Hispanic.” In contrast, only one of our patients fell into that class, yet our data were very similar to the Garway-Heath et al. data when both were plotted relative to mean normal (Figure 3). Our study extends the finding of linear relations between structure and function to a more diverse patient group.

Many studies have used correlations between perimetry and rim area, which is a weak method for assessing agreement²⁶. We performed a more complete analysis of agreement using the Bland-Altman method, and found that on average the two measures gave similar results at all levels of damage, once the measures were converted to fraction of mean normal (Figs. 3 & 4). The confidence limits from this analysis were similar for our data and for the Garway-Heath et al. data, showing scatter around the mean that was consistent with the variability found in control subjects. This means that, if perimetric sensitivity and rim area declined at the same rate, then normal variability in relations between the two could account for the scatter around the mean for our patients.¹⁵

We adapted the Garway-Heath map²⁰ relating visual field locations with optic nerve sectors, which was based on the 24-2 test locations. This required that we interpret that map for the 10-2 locations. We also developed five alternative maps (Fig. 1), from other maps in the literature and obtained the same results with all six maps, with only minor variations. Our results do not appear to depend on the particular map used for the temporal optic disc sector.

We did not find global disc area or age to be correlated with area of the temporal rim in normal eyes. Prior studies have found disc area to be a predictor of global rim area, with several different models for the relation between disc area and rim area^28–30. In Figures 2–4 we scaled the symbol sizes for our data by the disc area, and there was no obvious effect of disc area. We repeated our analyses after removing the three patients with the largest disc areas, and the results were unchanged.

The HRT-III uses refractive error to estimate the axial length of subjects, to account for magnification of retinal image size. Therefore differences between estimated and actual axial length can result in systematic errors in area of rim and disc. We did not take axial length measurements in our subjects, so this is a potential source of variability. Three-quarters of our patients has refractive errors within ±2 diopters spherical equivalent, so this source of variability should not be high. Figures 2–4 indicate refractive error by the grayscale of the symbols, and there was not an obvious effect of refractive error on our analysis.

We selected patients who had repeatable defects within 10° of fixation, which is typically considered to occur at moderate or advanced stages of disease. If “pre-perimetric glaucoma” were an earlier stage of disease, then the large majority of difference scores in Figure 3 would have been negative (corresponding to rim area being a smaller fraction of mean normal than perimetric sensitivity). However, only 15 of 31 values were negative. This result is consistent with the conclusion by Garway-Heath and colleagues⁶, who rejected the hypothesis of a functional reserve in the visual system. Our findings are in agreement with the analysis of Hood and Kardon^{15, 19}: agreement between structural and functional measures is strong, and is limited by normal between-subject variability. This results in some patients at an early stage of disease showing structural defects before perimetric defects, and other patients at the same stage of disease showing perimetric defects before structural defects, when defects are detected as statistically significant differences from mean normal values.

We gathered our own norms, rather than using machine norms. Our controls had perimetric sensitivities that averaged slightly lower than the machine norms for the 10-2, with a similar effect of age, and had mean global rim area that was slightly higher than machine norms for the HRT-II. If we had relied on machine norms, perimetric defects may have been on average more severe than rim defects.

None of our patients had rim area less than 20% of mean normal, even though a fifth had perimetric sensitivities of 6% to 17% of mean normal. This is reflected in Figure 3 as the absence of negative differences when mean is below 0.4, and is consistent with the presence of a non-neural component in structural measures. For RNFL thickness corresponding to inferior temporal and superior temporal optic disc sectors, Hood and Kardon proposed that the non-neural component averages 33% of mean normal, and data on blind eyes by Sihota³¹ are consistent with the non-neural component for the temporal sector being as large as for the other sectors. Recently spectral domain OCT has been used to assess thickness of the ganglion cell layer, and recent abstracts have compared this to results of perimetry with the 10-2 test locations^32–34; the results of these studies support the existence of a substantial non-neural component in this structural measure as well. We fit the Hood-Kardon model to our data and obtained values near zero for the non-neural component of rim area, and a one-tailed 95% confidence limit well below 0.33, so we conclude that the non-neural component is smaller for the neuroretimal rim area of the temporal sector than for RNFL thickness corresponding to inferior temporal and superior temporal optic disc sectors. Our study was not designed to address this question, so further research is warranted to assess this conclusion. If this finding is confirmed, then rim area may be more useful than RNFL thickness in assessing progression that can threaten fixation.

The goal in treating patients with glaucoma is to prevent or delay visual impairment, which is assessed in most clinical settings by perimetry. Perimetric defects within 10 degrees of fixation can have a profound influence on patients’ visual function and quality of life, yet for the typical 24-2 pattern the distribution of test locations in the central 10 degrees is limited to 12 locations, each separated by 6 degrees. The 10-2 pattern affords a less coarse measure of central vision function, as it samples 68 locations separated by 2 degrees in the central ten degrees of visual field. Prior studies of the relations between structure and function in glaucoma used the 24-2 pattern of visual field locations and included a majority of patients with early to moderate disease. The current study used the 10-2 pattern of locations and concentrated on patients with moderate to advanced disease in an effort to better understand glaucomatous defects that may have the greatest impact on patients.

The results of the present study support the notion that relations between clinical measures of structure and function in patients with glaucoma appear to be, on average, in general agreement. On an individual patient basis, however, there was often considerable disagreement between measures of structure and function. As this is a cross-sectional study, we are not able to determine whether, with time and progression of the disease, these measures will eventually approximate each other for each individual patient, nor whether the value of one measure is predictive of the other. This finding highlights the importance of longitudinal analyses of the relations between structural and functional measurements in the clinical management of patients with glaucoma.