Several studies have suggested that new imaging instruments are as accurate as stereophotography in diagnosing glaucoma. The patient population used in a study, however, is likely to be significantly different from that found in a clinical practice, and factors such as disease severity and the size of the optic nerve also will influence the diagnostic accuracy of these instruments outside of the clinical trial setting.
Several studies have found that diagnostic accuracy was similar when assessed with various imaging instruments such as optical coherence tomography (Stratus OCT, Carl Zeiss Meditec), confocal scanning laser ophthalmoscopy (HRT II, Heidelberg Engineering), and scanning laser polarimetry (GDx, Carl Zeiss Meditec) and also when compared with stereophotographs, he said.
These studies traditionally have been performed using a group of cases, usually glaucoma patients with confirmed visual field loss, and a group of controls, usually healthy individuals without any findings suggestive of disease. Diagnostic accuracy measures such as sensitivity and specificity are then obtained. This approach, however, may be poorly suited to the task of evaluating imaging instruments in clinical practice, Dr. Medeiros said.
In most studies, the typical patient has a clear glaucomatous visual field defect, whereas in a practice setting, there would be little point in using an imaging test to diagnose disease in such a patient, because the diagnosis already is established by the visual field defect, nor would the doctor apply the test to a healthy volunteer.
"The challenge is really the patient with suspected glaucoma," Dr. Medeiros said, noting as an example a patient who presents with a suggestive optic nerve appearance but with normal or unconfirmed visual field findings. "The bad news is that the population sample that is included in most studies may not be representative of the one in which we apply the diagnostic tests in everyday practice. Therefore, the estimates of sensitivity and specificity obtained from these studies may not be directly applicable to the clinical practice."
Dr. Medeiros, Robert N. Weinreb, MD, Distinguished Professor of Ophthalmology, University of California, San Diego, and colleagues recently evaluated this issue,1 studying the influence of the composition of the population on the diagnostic accuracy of one imaging instrument, the confocal scanning laser ophthalmoscope (CSLO). They first evaluated diagnostic accuracy to discriminate patients with visual field loss from healthy subjects. Then they conducted a second analysis to distinguish patients with suspected disease who had glaucoma from those who did not, which would be a more relevant question in the clinical setting.
Follow-up data were used in the second analysis to determine which of the patients with suspected glaucoma had the disease and which did not.
The difference in diagnostic accuracy between the two analyses was then reviewed, and significant differences were found, Dr. Medeiros said. In analysis one, the glaucoma probability score (an objective parameter provided by the CSLO) had an area under the receiving operating characteristic (ROC) curve of 0.89, which was comparable to that found in most studies that compare patients who have visual field loss with healthy subjects. In analysis two, the ROC curve area dropped to 0.65. Sensitivity dropped from 80% in the first analysis to 40% in the second.
"But we also have some good news," Dr. Medeiros said. "Longitudinal studies have shown significant value of the measurements obtained by imaging instruments in predicting who develops glaucoma and whose disease progresses. Also, a number of studies have predicted that imaging instruments may be able to detect preperimetric glaucoma. This suggests a potential for these imaging instruments to replace stereophotographs. However, we need to design studies of diagnostic tests that mimic the use of these tests in clinical practice. Studies using only subjects with well-defined visual field loss and healthy volunteers without suspected glaucoma will not accomplish this," he concluded.