This study has quantified the relationship between CDR and retinal ganglion cell number. Healthy eyes had a mean stereophotograph vertical CDR of 0.45 ± 0.15, which corresponded to a mean RGC count of 1,063,809 ± 174,254 cells (). The healthy eye RGC count estimate is consistent with the number of RGCs observed in previous histologic studies.34–37
For example, Jonas and colleagues34
found a mean optic nerve fiber count of 1,158,000 ± 222,000 in healthy eyes.
Our results showed that eyes with glaucoma had larger stereophotograph vertical and average CDRs, and significantly lower estimated numbers of RGCs than healthy eyes (, P
< 0.001 for all comparisons). The mean RGC estimate for glaucomatous eyes was 530,475 ± 267,241 cells. The stereophotograph vertical and average CDR increased with worsening severity of disease, and there was a corresponding decrease in RGC number, although the relationship between CDR and RGC count was not linear (). Similar results were found for Cirrus vertical CDR and Cirrus average CDR. A consequence of nonlinearity is that in eyes with a small CDR, loss of a relatively small number of RGCs could result in a large increase in CDR. In contrast, in eyes with a large CDR, the same loss of RGCs will result in a relatively smaller and perhaps undetectable change in CDR. This relationship has important implications for clinical practice, and suggests that CDR is a relatively insensitive measure of RGC loss in eyes with moderate to large CDRs. For example, it can be seen from that a change in CDR from 0.5 to 0.6 could occur with a loss of just approximately 44,000 RGCs. In contrast, it would take a loss of close to 300,000 RGCs for an eye to have a change in CDR from 0.9 to 1.0. Therefore, in an eye with advanced disease and large CDR, it will be difficult to detect further changes in CDR unless the loss of RGCs is substantial. These results may explain the widely held belief, also supported by some studies,2,20,38,39
that evaluation of the optic disc by biomicroscopy or photographs may be relatively insensitive to detect disease progression in moderate to advanced stages of damage.
The nonlinear relationship between RGC count and stereophotographic CDR may be explained by considering the relationship between CDR and neural tissue at the optic nerve head. The neuroretinal rim is composed of RGC axons, in addition to nonneuronal tissue. When RGC axons are lost there is also a reduction in neuroretinal rim area.34,40
Following a histologic study, Jonas and colleagues34
proposed that the relationship between neuroretinal rim area and the number of RGC axons is linear and that a 175,000 decrease in RGC axons is equivalent to approximately a 0.30 mm2
decrease in rim area.34
The neuroretinal rim may be considered as an annulus of tissue within the optic canal. Vertical CDR is a measure of neuroretinal rim in a single meridian and represents the ratio of the diameters of the inner and outer arc of the annulus. The relationship between the diameters of the annulus and the area of the annulus is described by the formula π(R2
), where R
is the radius of the outer circle of the annulus and r
is the radius of the inner circle, and is therefore nonlinear. An eye with a disc area of 2.0 mm2
will have an approximate radius of 0.8 mm. Therefore, if this eye has a CDR of 0.8, the annulus of neuroretinal rim would be π(0.82
− [0.8 × 0.8]2
) = 0.72 mm2
. With a CDR of 0.9, the calculated annulus of neuroretinal rim would be π(0.82
− [0.9 × 0.8]2
) = 0.38 mm2
. The loss of neuroretinal rim in this eye progressing from CDR of 0.8 to 0.9 would be estimated as 0.72 − 0.38 = 0.34 mm2
. Using the relationship of Jonas et al.34
of 175,000 RGCs per 0.30 mm2
rim area, a loss of 0.34 mm2
would correspond to 198,000 RGCs. This number is remarkably similar to that calculated by our model (), which shows that a change from CDR of 0.8 to 0.9 in an eye with 2.0 mm2
of disc area would correspond to a loss of 211,000 RGCs. Such agreement with histologic findings provides further validation to our analyses. Although the neuroretinal rim is not an annulus as the optic disc is usually vertically oval and rim loss is not concentric, the model supports the finding of a nonlinear relationship between CDR and RGC number.
Although CDR has been commonly used as a clinical measure for diagnosis and staging of glaucoma, the high intra- and interexaminer variability in gradings2,13–20
would suggest that relatively large CDR changes (0.2 or more) would be necessary for changes to be flagged as significant, especially if documentation is performed only by drawings or descriptions of the optic disc. Such changes would represent very large losses in RGC numbers, even in relatively early stages of disease. For example, a change in CDR from 0.5 to 0.7 would correspond to close to 130,000 RGCs (). In more advanced stages of damage, larger losses would be necessary. A 0.2 change in CDR from 0.7 to 0.9 would imply a loss of close to 350,000 RGCs. Such findings strongly imply that clinical assessment of CDR measurements should not be used as the sole method for detection of glaucomatous progression. In contrast to CDR changes, assessment of RNFL measurements with imaging technologies seems to be able to detect much smaller losses of RGCs over time.22,23
Interestingly, eyes with CDR of 1.0 still had an estimated RGC number close to 200,000 cells (). A similar “floor effect” has also been described with structural assessment by imaging technologies21,41,42
and it would indicate an inability of these methods to detect further progression once very advanced stages of damage have been reached. For CDR measurements, the reasons for such floor are likely to include preservation of neuroretinal rim in meridians not measured by CDR, the inability of examiners to discern small quantities of residual neural tissue or supporting elements, or artifact.
The shallow slope in the relationship between CDR and RGCs in eyes with small CDRs suggests that one might expect an increase in CDR with aging, particularly because aging is associated with a decrease in RGCs of 4000 to 8000 cells, or 0.3% to 0.5% per year.34–37,43
In clinical practice, significant enlargement of CDR with aging is usually not seen. This could be due to a lack of long-term studies investigating optic disc changes in healthy eyes. Alternatively, the lack of change in CDR with age-related loss of RGCs may be due to an increase in nonneural tissue in the neuroretinal rim. Previous studies have suggested that neural tissue losses are replaced by nonneural tissue and that axonal density decreases with aging.43,44
Because these different types of tissue cannot be distinguished by clinical exam, CDR changes would not be apparent. In addition, the shallow relationship between RGCs and vertical CDR in eyes with small CDRs may also be due to the variability of CDRs in normal eyes, with similar RGC counts. To investigate this possibility, we performed a separate analysis using only the glaucomatous eyes in the study but found a similar nonlinear relationship between RGCs and CDR.
Disc size was also a significant influence on the relationship between CDR and the number of RGCs (). There is a wide range of normal optic disc area, from smaller than 1 mm2
to as large as 5 mm2
in population-based studies using confocal ophthalmoscopy.45–47
Eyes with large optic discs have larger CDRs9
and also tend to have greater globe dimensions, larger neuroretinal rim area,48
larger retinal surface area, and a higher number of retinal photoreceptors and RPE cells.5,49
Disc area has also been proposed to influence glaucoma susceptibility.50,51
We found that eyes with large optic discs had a greater number of RGCs than eyes with smaller discs. The estimated number of RGCs increased by 165,000 for each 1 mm2
increase in disc area (). This estimate shows good agreement with the findings of a histologic study, which found that the number of RGC axons increased by 175,000 for a 1 mm2
increase in disc area.34
Given that eyes with large optic discs tend to have large CDRs, change in CDR may be a particularly insensitive measure of RGC loss in eyes with large discs.
We acknowledge the limitations of this study. Empirically derived formulae were used to estimate the number of RGCs and the original formulae were derived from studies in a primate model of glaucoma.21
Furthermore, the formulae were developed from studies using time-domain OCT and there may be differences in the RGC estimates obtained using SD-OCT. Despite these limitations, the formulae have been validated in multiple external human cohorts.22–24
Also, the RGC estimates we obtained were closely related to the findings of human histologic studies.34–37
In conclusion, the results of our study suggest that assessment of CDR is an insensitive method for evaluation of progressive neural losses in glaucoma. Even relatively small changes in CDR may be associated with large losses of RGCs, especially in eyes with large CDRs.