Sources of bias in studies of optic disc and retinal nerve fibre layer morphology

Interest in the morphology of the optic disc and nerve fibre layer in glaucoma has increased with the advent of new instruments able to make quantitative measurements of the optic disc and nerve fibre layer structure. Particular attention is given to the potential that these instruments may have to diagnose the condition(s) that we call glaucoma—that is, to classify individuals as “normal” or “glaucomatous”.

The usual way in which we evaluate a test is to apply it to a “normal”, reference, population and to a population believed to have the disease on the basis of other “gold standard” tests, and then count the proportions that are correctly classified in the two groups. In order to be able to generalise the results to clinical practice, the study population must reflect the population to which the test will be applied. Care, therefore, needs to be taken that restriction (entry) criteria applied to individuals do not bias the outcome. A frequent error is the use of an outcome variable of a test to restrict the entry of subjects into either the “normal” group or the “disease” group. In the context of glaucoma, we may want to know whether abnormal disc (or nerve fibre layer) morphology can correctly identify glaucoma patients. Restriction criteria should, therefore, not include optic disc appearance, but other gold standard tests, such as perimetry. Too often one finds a phrase such as “normal appearing optic disc” as a restriction criterion for the entry of subjects into a control group. Similarly, the phrase “optic disc cupping consistent with the diagnosis of glaucoma” is found as a restriction criterion for glaucomatous patients. The restrictions anticipate the outcome of the study, and make such studies impossible to evaluate. The results of such studies are merely a description of the groups selected to have a certain appearance of optic disc/nerve fibre layer, not of what is “normal” or “abnormal”. The findings cannot be extrapolated to the (unselected) general population.

To some extent, visual field restriction criteria may bias the findings of investigations into the pattern of glaucomatous morphological changes, because the topography of optic disc and visual field changes are likely to be related. Thus, selection of patients with “typical” glaucomatous field defects, such as the arcuate scotoma, will bias findings towards inferotemporal and superotemporal neuroretinal rim thinning. This bias is more difficult to avoid than that of direct selection on the basis of disc appearance. This is because glaucoma patients almost invariably come from the pool of clinic patients who have been “identified” as glaucomatous on the basis of currently recognisable signs. It may be difficult to overcome some of these biases, and these limitations should be recognised in the discussion of the results. Study designs that actively promote bias, such as the inclusion of optic disc morphology in the restriction criteria of studies investigating optic disc changes in glaucoma, should be avoided. Reviewers should ensure that the conclusions of such studies go no further than the study design allows.