Accurate and reliable clinical diagnosis is a cornerstone of Western Medicine. Reus et al. (699) investigate the accuracy of clinical diagnosis of glaucoma by performing an Europe-wide diagnostic test evaluation study. The study question was: Can general ophthalmologists diagnose glaucoma with certainty by looking at stereoscopic optic disc photographs alone, without access to any other data? The subjects were ophthalmologists, either randomly invited or 'hand picked' (we don't know the numbers of each) to participate in the study. The 'gold standard' diagnosis was an 'expert panel' from the Rotterdam Eye Hospital who probably (it is not mentioned) had access to all the relevant patient data such as visual field test results, IOP and other examination findings. The diagnostic accuracy of the GDx and HRT analyzers was also measured against this gold standard.

The ability of ophthalmologists to correctly diagnose glaucoma from a photograph only varies across Europe and machine analyzers do better than most

The median ophthalmologist could diagnose glaucoma as well (sensitivity) as the expert panel in three out of four cases, using only a photograph. Their ability to be certain a patient was normal based on the disc photograph (specificity) was 90%. The range of accuracy varied widely however, from 90% sensitivity and 100% specificity to one who picked 80% of the glaucomas as glaucoma but also over half the normals, to another who picked 80% of the normals' discs as normal, but also over half the glaucomas. The image analyzers performed better than some ophthalmologists, but not as well as others, depending on which test was used. But they all performed better than the median ophthalmologist. What can we learn from this study? First, the ability of ophthalmologists to correctly diagnose glaucoma from a photograph only varies across Europe and machine analyzers do better than most. But there is an implicit bias in this statement. Ophthalmologists are taught to diagnose people (rather than eyes) by taking a history and examining a live patient. So the ophthalmologists in this study were not given all the data they would usually use to make a diagnosis. In contrast, the image analyzers were given all the data they needed, including the patients age (which was not given to the ophthalmologists). Would this additional data have improved the ophthalmologists' results? We don't know. Nor do we know how representative these results were, as no data could be collected from the vast majority (over 70%) of those asked to participate. This study is nevertheless very important as is demonstrates the use of diagnostic test evaluation to improve clinical diagnosis, a very important area ripe for further research.