Patients and methods

The study included 102 “preperimetric” patients, 50 normal subjects, and 61 glaucoma patients (Table 1). All “preperimetric” glaucoma patients had an open chamber angle, increased IOP (measurements >21 mm Hg), an abnormal appearance of the optic nerve head, and normal visual fields examined by conventional computerised perimetry (Octopus programme G1). All patients in the glaucoma group had an open anterior chamber angle, increased intraocular pressure measurements above 21 mm Hg, an abnormal appearance of the optic disc, and glaucomatous visual field defects. An abnormal appearance of the optic nerve head was defined by the presence of an abnormal shape of the neuroretinal rim including rim notches, localised or diffuse loss of retinal nerve fibre layer, and disc haemorrhages. Mixed together with optic nerve head photographs of more than 500 other normal subjects and patients with non-glaucomatous or glaucomatous optic nerve damage, the optic disc slides were evaluated in a masked fashion by two examiners (WMB and JBJ) without knowledge of the visual field data or IOP measurements. One of the examiners had experience in morphological evaluation of the optic nerve heads of more than 500 patients, the other examiner has been active in this field for more than 10 years. If uncertainty existed in the clinical categorisation of the patients, which was based on one single cross section assessment, the more experienced examiner decided upon the diagnosis. For the evaluation of the retinal nerve fibre layer, black and white wide angle photographs of the retinal nerve fibre layer were additionally available for all patients included in the study.

A glaucomatous visual field defect was defined as a Octopus G1 field with (a) at least three adjacent test points having a deviation of equal or greater than 5 dB and with one test point with a deviation greater than 10 dB lower, (b) at least two adjacent test points with a deviation equal to or greater than 10 dB, (c) at least three adjacent test points with a deviation equal to or greater than 5 dB abutting the nasal horizontal meridian, and (d) a mean visual field defect of more than 2 dB.

The glaucoma group was differentiated into patients with primary open angle glaucoma in whom no obvious reason for the elevation of IOP had been detected (n=38) and into patients with secondary open angle glaucoma due to pseudoexfoliation of the lens (n=5) or primary pigment dispersion (n=18) (Table 1). To avoid bias by small and large optic discs, disc area was limited to 1.8–2.8 mm².

Patients with a myopic refractive error exceeding −8 dioptres were excluded owing to different morphology of the optic disc.10 If both eyes were examined, only one randomly selected eye per subject was taken into for statistical analysis.

For each eye, three 10° confocal scanning laser tomographic images were obtained using the Heidelberg retina tomograph (HRT) (Heidelberg Engineering, Heidelberg, Germany, software version 2.01).11 12 The mean topography of the three images was analysed by the HRT software version 2.01, as proposed by Burk and co-workers.12 Mean deviation of the images used in this study was 24.1 (SD 7.8) μm. The border of the disc was outlined manually with the help of the optic disc photographs projected simultaneously. According to software 2.01, reference plane for the delineation of the optic cup from the neuroretinal rim was the level 50 μm beneath the contour line in the temporal segment, 4–10 degrees below the horizontal axis of the disc. We determined the HRT variables disc area, volume, and area of the neuroretinal rim (volume/area above reference plane) and volume and area of the optic cup (volume/area below reference plane), cup shape measure (third central moment of the frequency distribution of depth values, relative to the curved surface of all parts in the contour line the steeper the cup walls, the more positive the value), and thickness of the retinal nerve fibre layer (measured along the contour line). The HRT variables were measured for the optic disc as a whole, in a right angled superior sector and a right angled inferior sector. The middle lines of both sectors were tilted 13° temporal to the vertical optic disc axis. For all eyes, 15° colour stereo optic disc transparencies were additionally taken.13 HRT examination and optic disc photography were performed on the same day.

Results

All tested HRT variables were significantly (p<0.05; ranked Mann–Whitney U test) different between the normal group and the “preperimetric” glaucoma group, with the exception of the cup shape measure, measured globally and separately in the inferior disc sector, rim area in the inferior sector, cup area and volume in the inferior sector and global cup (Table 2). Parallel to marked interindividual variability in each group, there was pronounced overlap in all measured optic disc variables between the two study groups (Table 2) (Figs1, 2, 3).

To differentiate between the normal eyes and the preperimetric glaucoma eyes, neuroretinal rim area in the superior sector (24.8%), thickness of the retinal nerve fibre layer in the inferior sector (26.5%), and rim volume above the reference level in the inferior sector (25.5%) showed highest sensitivity values at 95% specificity (Table3).

Comparing the normal eyes with the glaucoma eyes with perimetric defects, all HRT variables were significantly (p<0.001) different, again with marked overlap between the study groups (Table 2) (Figs1, 2, 3). Neuroretinal rim area in the superior sector (72.1%), rim volume above the reference plane global (67.2%), and in the superior sector (73.8%) showed highest sensitivity values at 95% specificity. The cup shape measure in the inferior sector (26.2%) showed lowest sensitivity (Table 3).

For further separation between the normal eyes and the glaucoma eyes with perimetric defects, a discriminant function was derived for multivariate presentation. The function coefficients of the present study are given in the formula of the value (D) of the discriminant function: D= −2.77 + 0.3 rim area + 3.7 rim volume + 4.3 retinal nerve fibre layer −3.7 cup shape −3.1 cup volume −0.9 cup area. This score, including six HRT variables, was used to evaluate sensitivity and specificity with receiver operator characteristics curve. The overall score (D) including all six measurements reached a sensitivity of 83.6 % at a specificity of 95%. It means that the combination of the six optic disc variables had a higher diagnostic power than each of the variables taken alone. For the differentiation of the normal eyes and eyes with preperimetric glaucoma, the same multivariate approach increased the sensitivity value to 42.2% at 95% specificity (Table 3).

Discussion

Traditionally glaucoma has been defined by the triad of increased IOP, optic disc changes, and visual field defects. Histological studies have shown however that there can be a significant loss of ganglion cells before evidence of functional loss on conventional achromatic visual field testing.14 15 For this reason attention has been focused on alternative, more sensitive ways of detecting early ganglion cell damage than it is possible with white on white perimetry. In several studies on eyes with elevated intraocular pressure and normal visual fields, abnormal results in various psychophysical and electrophysical examinations were reported.16-21 As with the newer psychophysical and electrophysiological techniques, it has been shown that abnormalities in the appearance of the optic disc may precede visual field defects.4-9 22 23 These abnormalities include an unusually small area of the neuroretinal rim, an abnormal shape of the rim, high cup to disc ratios, an usually large parapapillary chorioretinal atrophy, a decreased visibility of the retinal nerve fibre layer including localised defects, and the presence of splinter-shaped haemorrhages at the optic disc border. The variables with the highest predictive value to differentiate between normal eyes and eyes with increased IOP and normal visual fields were shape of the neuroretinal rim and visibility of the retinal nerve fibre layer.9 Presence of localised defects of the retinal nerve fibre layer and splinter-shaped disc haemorrhages had high specificities; however, their sensitivity was relatively low.

Most of these variables are qualitative or semiquantitative. They can be evaluated by the ophthalmoscopy without the necessity of measuring them. In an automated analysis of the optic nerve head, however, quantitative variables are determined such as area and volume of the neuroretinal rim and depth of the optic disc cup. With confocal laser scanning tomography as a technique for computerised analysis of the optic nerve head becoming more popular,11 12 24-28 the purpose of the present study was to evaluate sensitivity and specificity of quantitative optic disc variables in the differentiation between normal eyes and eyes with increased IOP and normal visual fields.

In the normal group, we found a pronounced interindividual variability for all optic disc variables measured (Table 2) (Figs1, 2, 3). Similar results have been obtained when the optic disc variables were measured by planimetry of stereo optic disc photographs.13 29 The interindividual variability of the HRT data may be caused by the heterogeneity of a biological group, the influence of methodological factors such as varying conditions in taking the HRT images, or the observer associated variability in defining the disc margin, or by other factors. The marked interindividiual variability being typical also for many other quantitative biological variables such as body height and weight may be the reason why the normal group and the preperimetric glaucoma group showed a pronounced overlap in the quantitative optic disc variables (Figs 1, 2, 3). Correspondingly, sensitivity and specificity of these variables were relatively low to differentiate between the normal eyes and the eyes with preperimetric glaucoma (Table 3) (Figs 1, 2, 3, Fig 4). To cite an example, the optic disc variable rim volume above reference superior detected glaucomatous optic nerve damage in 26% of the preperimetric eyes at a given specificity of 95%, although all of these eyes showed glaucomatous optic nerve damage as described by qualitative optic disc variables. It shows that, by taking only one quantitative optic disc variable, it is rather difficult to detect early glaucomatous optic nerve damage. It suggests that the combination of several quantitative disc variables in the form of a discriminant analysis should increase the likelihood of detecting glaucomatous abnormalities in optic nerve heads which, by ophthalmoscopic evaluation, show qualitative abnormalities such as presence of localised retinal nerve fibre defects. In the present study discriminant analysis increased sensitivity in the preperimetric group from 27% for the best single variable (RNF thickness inferior) up to 42.2%. In the group of perimetric glaucomatous optic discs sensitivity at 95% specificity increased from 73.8% for the single variable rim volume above reference superior to 83.6% by multivariate analysis.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Box plots showing the distribution of neuroretinal rim area in three groups: controls, preperimetric open angle glaucoma (prePOAG), and perimetric open angle glaucoma (POAG).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Box plots showing the distribution of neuroretinal rim volume above the reference level in three groups: controls, preperimetric open angle glaucoma (prePOAG), and perimetric open angle glaucoma (POAG).  

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Box plots showing the distribution of retinal nerve fibre (RNF) layer thickness in three groups: controls, preperimetric open angle glaucoma (prePOAG), and perimetric open angle glaucoma (POAG).

• Download figure
• Open in new tab
• Download powerpoint

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

ROC (receiver operated characteristic) curves generated for multivariate analysis (broken line) and optic disc variables measured by confocal laser scanning tomography in the preperimetric and perimetric glaucoma group. (A) Variables of the rim: rim area superior (more broken line), rim volume superior (solid line), RNF inferior (dotted line). (B) Variables of the cup: cup volume superior (broken line), cup area superior (solid line), cup shape measure (dotted line).

The variables area and volume of the optic rim and cup and cup shape measure showed higher sensitivities in the superior than in the inferior sectors both in the preperimetric and the perimetric group at 95% specificity. These findings seem to stand in contrast with earlier reports showing pronounced rim loss in the inferior sector of the optic disc and perimetric field loss in the superior half of the visual filed in early glaucoma.22 30-32 As pointed out earlier, the retinal vessels are included into the measurements of cup and disc by the HRT.27 One could speculate that the proportion of the vessels contributing to rim and cup volume is smaller in the superior sector, as the disc is scanned nearly perpendicularly and the values for the HRT variables are more valid.

The results of the present study confirm previous investigations in which confocal scanning laser tomography was used to measure quantitative optic disc variables in eyes with early glaucoma.26 28 Defining early glaucomatous visual field loss as mean perimetric defects of less than 5 dB, Mikelberg and co-workers26 reported highest sensitivity at 89% and 78% specificity using the papillomacular bundle as a reference plane and cup shape measure being one of the most informative variables. Ushidaet al  28 found 83% sensitivity at 86% specificity with the variable cup shape measure. In the present study, sensitivity and specificity values were lower than in the two other studies cited. The reason may be that glaucomatous optic nerve damage was less marked in the preperimetric group of the present study than in the groups of the investigations previously cited in which the glaucomatous eyes had manifest perimetric defects. As in the present study, Mikelberg and others26 28 33 also found that a combination of optic disc variables in the form of a multivariate approach improves the discrimination between normal and glaucomatous eyes.

The present cross sectional study did not evaluate the ability of the HRT in detecting change in follow up examination, which has already clearly been demonstrated by Hatch and co-workers.34Future studies may reveal whether quantitative data of the optic disc and retinal nerve fibre layer as measured by other techniques such as planimetry of the optic disc photographs,13 29polarimetry,35 or optical coherence tomography36 have a higher diagnostic power than optic disc variables determined by confocal scanning laser tomography, or whether quantitative optic disc variables are generally more useful for follow up examination of patients34 than for the detection of glaucomatous optic nerve damage in a cross sectional evaluation.