The classic teaching is that when assessing a glaucoma patient we should look for a concordance between structural changes in the optic disc and functional changes as detected by perimetry. When this is identified, such as in the case of focal loss of neuroretinal rim (e.g. an inferior notch) combined with a well-defined scotoma (e.g. a superior arcuate defect), we are reassured that what we are looking at is definitely glaucoma, and it makes sense with regard to our understanding of the pathological process. If there is a mismatch then we should consider other diagnoses. Of course the site of primary damage is still in debate, but loss of a bundle of nerve fibres and death of the corresponding retinal ganglion cells will typically produce a nicely defined scotoma that matches the topography of the retinal nerve fibres. However, we know in the clinic that it does not always work this way, and we are frequently confounded by our interpretation of discs where the structural appearance and the functional loss do not match as well as we might have expected. The recent addition of high resolution imaging devices to assess the retinal nerve fibre layer (RNFL) in vivo has given us potentially a more accurate means to document structural change, but the structure/function relationship demonstrated by these devices has not been overwhelming. There are a number of reasons for this, relating to variations in anatomy and function among the population, variations in methods of assessment and the different measuring and scaling systems used to quantify changes. In this issue, Malik et al. review in detail the factors that define this structure/function relationship and re-examine the literature to explain why there are discrepancies. They go on to challenge the very frequently quoted notion that at least 25% of ganglion cells are lost before any functional change is evident.

A major obstacle in defining the relationship is the huge variety of optic disc sizes and shapes in the community, and this is associated with variations in the number and the distribution of nerve fibres around the disc margin. In fact, even our ability to actually define the disc margin is coming into question. Reis et al.1 and Strouthidis et al.2 have just reported that varying combinations of the termination of Bruch's membrane, border tissue or the anterior scleral canal opening produce the two-dimensional disc margin we see and identify. Photographs and confocal scanning laser tomography examinations may even overestimate the amount of remaining rim in regions where the full extent of Bruch's membrane is not detected clinically. Spectral domain optical coherence tomography can demonstrate transparent overhanging Bruch in some areas, and the angle of insertion of the nerve can make a significant difference to the perceived neural canal opening. In addition, when interpreting an obliquely inserting disc, the typical double hump pattern characteristic of a ‘normal temporal-superior-nasal-inferior-temporal (TSNIT)’ distribution gets displaced. The superotemporal and inferotemporal bundles are still usually within normal limits in terms of number of fibres but they will not fit the standard distribution of sectors represented in the normative database for a particular imaging device. In fact, nerve fibre layer studies show considerable variability for a given visual field location,3 with optic nerve head entry points varying with a standard deviation of 8.8°, with some individual points being spread over a range of up to 30°. Furthermore, a small but significant number of individuals do not have the anatomically well-defined horizontal raphé that we take for granted when interpreting fields, with some having overlaps that can produce variations in the typical nasal step pattern.4 Ideally, we need to individualize our structure/function map for each subject, which would need to take into consideration the size and three-dimensional shape of the disc, as well as related retinal anatomy, such as the fovea to disc angle, and position and/or overlap of the raphé.

There may be a variety of processes occurring in the spectrum of glaucomas, and evolution of the cupping process may vary considerably between individuals. A focally notched disc shows quite a different structural change from a more saucerized sclerotic type disc. It is assumed that the amount of nerve fibre loss should still correlate with field loss, but measuring this could be complex with factors such as lamina structure playing a role. Primate modelling of glaucoma has even shown some initial thickening of pre-laminar tissue in early cases, prior to the deepening of the cup. There is also a residual non-neural component of tissue (around 30–60 µm) that remains even in advanced glaucoma and will be included in the RNFL measured by our imaging devices, confounding the relationship.

The different methods employed by our test strategies for perimetry and imaging provide an inherent problem in combining data from the two sources. The structure function topographical map described by Garway-Heath et al.5 is the most commonly used, although more recently other groups have devised alternate models with slightly different representations.6 Such maps provide a representation of areas of the visual field that correlate with sectors of the neuroretinal rim. However, as RNFL is measured on a linear scale and visual fields on a logarithmic scale, the relationship between the two will not be linear. A small dB change in thresholds represents a much greater order of magnitude of change than the associated RNFL changes measured on a linear scale in microns.

Clinical trials have shown that there are many patients who show structural change before functional loss, but equally there are other studies that show a large number of converters on perimetry before disc end points are detected. This can largely be explained by our lack of ability to reliably define and detect change and the particular definitions used in these studies. Our field strategies use the same stimulus at all eccentricities even though we know that retinal cell distributions change with eccentricity. Electrophysiology studies with the multifocal visual evoked potential (VEP) have shown amplitudes to have a linear relationship with RNFL thickness and this strategy uses different stimulus sizes with increasing eccentricity (although mainly due to cortical magnification factors).

Typically, in early glaucoma, structural loss appears greater than functional loss, whereas in more advanced disease, it appears as if function changes at a greater rate than structure. However, if we could measure individual ganglion cell counts and had more sensitive techniques to assess early functional loss, this might not be the case, and the two would run in parallel. Ganglion cell dysfunction prior to actual death of the cell may account for some cases where perimetric loss appears to occur first.

Recent studies have suggested that combining data, for example using Bayesian hierarchical models,7 better predicts future progression. It makes good sense to combine all information when diagnosing a patient, and it appears that technology is moving towards trying to help us, with new printouts combining structural and functional data. It will probably take some time before we see the concordance we would like to be certain in our diagnostic process, but in the mean time we should accept that there will be some spurious and even confusing results that will require our clinical judgment to interpret.


  1. Top of page
  2. References
  • 1
    Reis ASC, Sharpe GP, Yang H et al. Optic disc margin anatomy in patients with glaucoma and normal controls with spectral domain optical coherence tomography. Ophthalmology 2012; 119: 73847.
  • 2
    Strouthidis N, Yang H, Reynaud JF et al. Comparison of clinical and spectral domain optical coherence tomography optic disc margin. Invest Ophthalmol Vis Sci 2009; 50: 470918.
  • 3
    Jansonius NM, Nevalainen J, Selig B et al. A mathematical description of nerve fiber bundle trajectories and their variability in the human retina. Vision Res 2009; 49: 215763.
  • 4
    Jeoung JW, Kim TW, Kang KB et al. Overlapping of retinal nerve fibers in the horizontal plane. Invest Ophthalmol Vis Sci 2008; 49: 17537.
  • 5
    Garway-Heath DF, Caprioli J, Fitzke FW et al. Scaling the hill of vision: the physiological relationship between light sensitivity and ganglion cell numbers. Invest Ophthalmol Vis Sci 2000; 41: 177482.
  • 6
    Kanamori A, Naka M, Nagai-Kusuhara A et al. Regional relationship between retinal nerve fiber layer thickness and corresponding visual field sensitivity in glaucomatous eyes. Arch Ophthalmol 2008; 126: 15006.
  • 7
    Medeiros FA, Zangwill LM, Girkin CA et al. Combining structural and functional measurements to improve estimates of rates of glaucomatous progression. Am J Ophthalmol 2012; 153: 1197205. e1.