advertisement

Editors Selection IGR 11-3

Visual Field: RNFL relationship and VF mean sensitivity

Comment by Paul Artes on:

24138 Structure-function relationship depends on glaucoma severity, Gonzalez-Hernandez M; Pablo LE; Armas-Dominguez K et al., British Journal of Ophthalmology, 2009; 93: 1195-1199

Find related abstracts

The link between structure and function in glaucoma has puzzled clinicians and scientists for a long time. It bears on which tests are most useful for patients with the disease and for those at risk of developing it, and how clinical data from different instruments can be interpreted and related to each other. These are important questions. Previous studies have shown how structure-function relationships depend on scaling ‐ they are typically curved when function is expressed logarithmically (e.g., dB, logMAR) and structure linearly (e.g., retinal nerve fibre layer thickness, RNFL, microns) and more nearly linear when the measures are expressed both either in logarithmic or in linear form. Equally important, but more controversial, is how scaling bears on the variation between, and within, individuals ‐ i.e., the predictability of visual function from ocular structure, and vice versa.

In a recent paper, Gonzalez-Hernandez et al. (1068) examine the relationship between RNFL thickness and visual field mean sensitivity across a large spectrum of glaucomatous damage. RNFL thickness was derived from the HRT - but unlike OCT, confocal scanning laser tomography does not measure the RNFL thickness directly but rather derives it from the surface topography of the optic nerve head. The negative RNFL thicknesses of some eyes ought to have raised concerns, or at least merited a comment.

The data illustrate a familiar dilemma: the evidence for a relationship between the estimates of structure and function, be it linear or curved, is hardly convincing. But how much of the scatter is due to the techniques applied to measure structure and function, and how much can be attributed to real physiological discrepancy? Perimetry and imaging technologies have evolved chiefly to obtain high diagnostic accuracy, and not to obtain high structure-function correlations. Much of the nerve fiber layer, for example that of the macula, is rather poorly represented by few test locations in the visual field, and if one confines the investigation to global rather than local measures one is likely to 'average out' some interesting detail.

The paper states that the linear correlation between structure and function was the same (to three decimal places) when the visual field was summarized in linear and logarithmic units. This strikes me as unlikely and is probably an error, but the data (Fig. 2) show that no transformation would materially alter the lack of a close link between the variables.

Despite the poor relationship between what is believed to represent RNFL thickness and visual field mean sensitivity (Fig. 2), the authors write that in patients with little or no visual field damage, 'function can be accurately estimated using morphological parameters..,' which is not really the case at all, neither in patients with good nor in those with bad visual fields. Figure 4 shows that there is no relationship between the predicted and observed visual field mean sensitivities. Looking for differences in the structure-function relationship across the spectrum of functional damage, the authors split their data into two groups, based on the function variable (mean sensitivity). This is not a good way of doing a subgroup analysis ‐ splitting the data on either variable of interest will almost always weaken the relationship and bias any gradient towards zero. A useful thought-experiment is the relationship between height and weight, and what would happen if one determined it separately in short and tall people. A better approach would have been to split the dataset depending on diagnosis, for example.

The authors are probably correct in concluding that the relationship between structure and function has something to do with the large physiological variation of ocular structure, and that it is difficult to determine functional damage from structural data. I am less convinced about 'limitations in the dynamic range of the morphological tests' ‐ particularly when the sub-zero RNFL thicknesses in their data suggest that the dynamic range is somewhat larger than it ought to be. Given that the relationship between RNFL thickness and visual field sensitivity has already begun to be investigated using more appropriate techniques, topographically resolved data, and arguably better numerical analyses, this paper adds little to our knowledge of structure and function in glaucoma.

• ← Previous
• Next →