Results

The study population characteristics are summarised in Table 1.

Multiple linear regression analysis of the planimetry data from the control group demonstrated a strong relation between the vertical cup diameter and the vertical optic disc diameter (T = 8.14, p <0.000 ) and a weaker relation with age (T = 2.04, p = 0.045), theR ² value for the regression was 0.45, p <0.000. The equation for the regression line is

vertical cup diameter =−1.537 + (1.204 × DD) + (0.0037 × age)

Optic cup diameter was not found to be related to sex, eye side, intraocular pressure, refraction, or keratometry. Optic disc diameter was not related to age.

The relation between cup and disc diameter is shown graphically in Figure 2. Omitting age as a variable, linear regression gives the following equation (R ² = 0.42, p <0.000):

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Plot of vertical cup diameter against vertical disc diameter. Mean (regression line) (R² = 0.42, p <0.000) and 95% prediction intervals shown.

vertical cup diameter = −1.31 + (1.194 × DD)

The upper limit of normal (95% prediction intervals) is expressed by

vertical cup diameter = −0.87 + (1.193 × DD)

The vertical CDR (Fig 3) may, therefore, be expressed by

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Plot of vertical cup/disc ratio against vertical disc diameter. (A) Control subjects. (B) Glaucoma patients. Solid line, mean (disc size dependent); broken line, upper limit of normal (disc size dependent); dotted line, upper limit of normal (disc size independent).(The vertical disc diameter is calculated from measurements of optic disc photographs using the Littmann correction for ocular magnification. When measuring the optic disc at the slit lamp, the correction factor for the 78 D lens is 1.11 and for the 90 D lens 1.39 (unpublished data presented at Glaucoma Society (UK & Eire) Annual Meeting, November 1996)).

vertical CDR = (−1.31 + (1.194 × DD))/ DD

and the upper limit of normal (95% prediction intervals), by

vertical CDR = (−0.87 + (1.193 × DD)) /DD

Using this equation, the sensitivity and specificity of the CDR to identify glaucomatous discs was 62.3% and 98.9% respectively. This compares with the cut off of the upper 97.5 percentile (0.682) which gives a sensitivity of 56.6% and specificity of 97.7%. The sensitivity and specificity for different cut off values of CDR is shown in Table 2. The sensitivity to identify discs as glaucomatous varies according to disc size. The two methods are compared in three disc size groups (small, lower quartile; medium, middle quartiles; large, upper quartile) and the results are given in Table 3.

The normal ranges for neuroretinal rim area can be defined by the prediction intervals of the linear regression between neuroretinal rim area and optic disc area, for the whole disc and each 30 degree segment.28 In this way, rim loss can be characterised as diffuse (neuroretinal rim area for the whole disc outside the normal range) or focal (segments only outside the normal range). Of the glaucoma patients, 11 of 53 had diffuse rim loss; 10 of these had a CDR outside the normal range on the disc size dependent measure. In addition, 23 of 42 patients with focal loss had an abnormal CDR.

Agreement between clinical (observer 1) and planimetric estimates for CDR in the ocular hypertensive subjects is shown in Figure 4. The mean difference was 0.03 with a standard deviation of 0.07. There was a slight, but significant, tendency for the clinician to underestimate small cups relative to the planimetric measurement (R ² = 0.15, p = 0.003).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Agreement between clinical and planimetric estimation of cup/disc ratio (ocular hypertensive patients). Mean and 95% prediction intervals shown.

Agreement of clinical estimates between the two observers was excellent. Ninety three per cent of estimations were at or within 0.05 CDR and differences of >0.05 occurred at small values of CDR, where the CDR has only a small affect on neuroretinal rim area. There was a slight, but significant (R ² = 0.13, p = 0.05), tendency for observer 1 to underestimate small cups and overestimate large cups, relative to observer 2 (Fig 5).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Agreement between two clinicians’ estimation of CDR (ocular hypertensive patients). Mean and 95% prediction intervals shown.

Discussion

Since the last century enlargement of the optic cup has been recognised as an important morphological feature of the glaucomatous eye.29 Cupping is an early sign and progression of cupping has been observed in ocular hypertensive patients6 30 31and in glaucoma patients, before the onset of visual field damage.27 32-34 An enlarged cup is a predictor of subsequent field loss.34-37

In attempting to quantify the size of the cup, many observers have compared the size of the cup to the size of the disc—as a ratio of cup area to disc area,38 39 cup circumference to disc circumference,40 or cup diameter to disc diameter.14 41-43 The simplest method, for the clinician examining a patient at the slit lamp, is to express cup size as the ratio of cup diameter to disc diameter. Initially, this was estimated in the horizontal meridian,14 42 but since it has been emphasised that the cup tends to enlarge in the vertical meridian in glaucoma,10 44-46 the estimate is now usually made in the vertical meridian.11 15 31

It has been recognised for some time that there is a wide range of optic cup size in normal eyes,14 15 38 42 47 with CDRs from 0.00 to 0.87 at the extremes.2 Much of the variability in cup size results from the physiological relation between the size of the cup and the size of the optic disc. This relation was rediscovered in the 1970s,2 16 17 although it was demonstrated by Elschnig on the basis of histological work at the end of the last century.47 The rate at which cup diameter increases with disc diameter, as defined by planimetry in this study, is very close to that found by planimetry in a population based study from Sweden: vertical cup diameter = −1.405 + (1.16 × vertical disc diameter) + (0.002 × age).48 This relation means that the CDR also varies with disc size.17 Healey and colleagues recently reported a population based study in which they determined the increase in CDR with increasing vertical disc diameter.49 The mean CDR found was 0.43, which compares with 0.44 in this study. Exact comparison of optic disc size is not possible, as a different method was used to correct for ocular magnification23 and camera magnification is not stated. However, they report an increase in the CDR from 0.26 to 0.55 when disc size increases from 1.00 to 2.00 mm. A similar increase in CDR was found in this study as disc size increased from 1.40 to 2.10 mm (Fig3). Healey and colleagues concluded that the change in CDR with disc size was of sufficient magnitude to warrant the measurement of disc size when assessing the CDR.

Initial attempts to find a useful CDR cut off value to separate normal and glaucomatous eyes were made without the knowledge of the relation between cup size and disc size and on the basis of the distribution of absolute values for CDR in the normal population. In a number of studies, the proportion of normal subjects with a CDR of 0.65 or greater ranged from about 2.2% to 4%14 15 35 and a study of ocular hypertensive and glaucoma patients50 found that the incidence of visual field defects increased markedly with CDRs greater than 0.7. However, the ability of the CDR to distinguish between normal and early glaucomatous optic disc has been found to be poor with a sensitivity and specificity of 64% and 57% from a clinic based study51 and 52% and 89%, at a cut off of ⩾0.5, from a population based study.52 A cut off of ⩾0.5 in this study gives a specificity of 64% and sensitivity of 92% (Table2). Calculations of sensitivity and specificity, for any given cut off value, depend entirely on the nature of the groups compared, and it is interesting to consider how these groups might differ to give such disparate sensitivities and specificities. It has been established that the normal optic disc morphology may vary between racial groups.53-55 Black Americans have larger, more vertically oval discs, with a larger vertical CDR.53 55 Thus, in a mixed population, a cut off of ⩾0.5 will have lower specificity, with a similar (or slightly greater) sensitivity. It may be possible to overcome this difficulty by using the nomogram presented in this paper (Fig 3), which takes disc size differences into account. The nomogram was derived from our control group, which comprised only white subjects, and the application of the nomogram will only be valid if the relation between optic disc size and CDR is similar in racial groups other than white. This needs further investigation.

Evaluating a test such as this on a glaucoma clinic population may result in overestimation of the sensitivity of the test. Although optic disc cupping was not assessed for inclusion of patients into our early glaucoma group, the patients were largely taken from the glaucoma clinic where the diagnosis of glaucoma is made, in part, on the basis of cup enlargement. Application of the method to a population based group would give a better indication of the true sensitivity of the test.

Variation between observers in the judgment of the CDR is notoriously high.56 However, it has been clearly demonstrated that agreement can be substantial given the right conditions.12 13 Agreement between observers,13 56 and single observer consistency,56 is improved under binocular (stereoscopic) conditions. Differences in observations occur when monoscopic and stereoscopic conditions are compared. Binocular viewing of the disc often results in a CDR estimation larger than that under monocular conditions.13 15 56 The effect of examination methods, and of the definition of the variables being measured, on observer agreement has been emphasised.57 58 The interobserver variation is higher, and agreement worse, in studies in which the participants are not given a definition of the variables to be defined,56 than those in which the variables are defined,13 and those in which the participants have undergone training in standardisation.12 It is for these reasons that both viewing conditions and the definition of the variables to be defined must be standardised. Differences in variable definition may also account for the differences found between clinical estimates of the CDR and those derived from image analysers.

A number of techniques to improve the accuracy of clinical cup size estimation have been proposed. These range from careful drawings,38 comparison with diagrammatic charts,14 39 41 photography with subsequent measurement,14 43 59 to measurement of the aerial image of the disc using the indirect ophthalmoscope and calipers.9 The simplest method, for the clinician examining a patient at the slit lamp, is to express cup size as the ratio of cup to disc diameter by comparison with a chart. Whether this results in acceptable accuracy and interobserver agreement needs to be assessed, but the agreement between the two observers in this study suggests that accurate and reproducible assessments are possible.

In this study we found a small, but significant, increase of cup size with age. We have discussed this finding in detail in another publication.60

The modest improvement in sensitivity and specificity in the whole group, obtained by taking optic disc size into account, does not reflect the greater advantage to be gained in clinical practice. The characteristics of patients referred to a glaucoma clinic are different from those of the general population. Many are referred with large cups and normal visual fields, a significant proportion of which will be individuals with physiological cups in large optic discs, and many are referred with moderate cups and risk factors for glaucoma. Subjects with normal sized cups and no risk factors are not referred.

Inspection of Table 3 shows that the ability of the disc size sensitive measure to detect glaucoma cases is greater in smaller (12/15) than in larger (3/8) optic discs. The explanation for this is that loss of neuroretinal rim tissue has a greater impact on the CDR when the CDR is small than when it is large.61 In our data set, a disc with a vertical diameter of 1.40 mm will have a CDR of 0.26. Loss of 10% of the rim area will change the CDR to 0.40, a difference of 0.14. A disc with vertical diameter of 2.10 mm will have a CDR of 0.57. Loss of 10% of the rim area will change the CDR to 0.63, a difference of only 0.06.

A disc size independent cut off will identify all CDRs above a certain level as abnormal. The sensitivity to detect glaucoma will thus be high in large discs, but at the cost of a much increased false positive detection rate in normal, but large, discs.62 This is clearly demonstrated in Figure 3: as disc size increases, more false positive diagnoses are expected if a single CDR cut off value is used.

In clinical practice, when assessing glaucoma suspect referrals with large optic cups but no risk factors for glaucoma, confidence that the optic disc may be normal will be greatly increased by a knowledge of disc size. In referrals with risk factors, knowledge that the disc is large should alert the clinician to exercise caution in placing too much reliance on the size of the cup alone when arriving at a diagnosis. Conversely, in referrals with risk factors and moderate cupping, knowledge that the disc size is small will enable the clinician to suspect glaucomatous damage.

Assessment of the relative size of the optic cup is only part of the clinical evaluation of the optic disc in glaucoma. Other important aspects include assessment of the shape of the neuroretinal rim,2 status of the nerve fibre layer,63 and site and extent of parapapillary atrophy.64 65 In this study there was a high glaucoma detection rate in eyes with diffuse neuroretinal rim loss using the CDR with respect to disc size. Those with notching of the rim and other signs may be picked up by qualitative assessment of the disc.

The actual relation between disc size and CDR, with confidence intervals, in a defined population has not previously been published, and it has therefore not been possible to make use of the knowledge that disc and cup size are related. The nomogram presented in this paper should be a substantial aid to the clinician. The value of the measurement of the CDR in relation to optic disc size is as an important adjunct to the qualitative assessment of the optic disc of glaucoma suspects.