advertisement
Reproducibility was particularly high for discs that were determined to show no change. Not surprisingly, the problem area was the confirmation of discs suspected to show deterioration. A measure of this ability to consistently define change, sensitivity, was only 64% to 81%Parrish et al. (796) address the reproducibility of graders to detect glaucomatous optic nerve damage from serial disc photographs. Follow-up photographs were compared with baseline by two masked graders; disagreement between the graders was adjudicated by a third grader. Optic disc deterioration was only determined after a confirmatory photograph was judged to exhibit change in suspected cases. The kappa statistic, a measure of agreement, was in the good to excellent range overall. Reproducibility was particularly high for discs that were determined to show no change. Not surprisingly, the problem area was the confirmation of discs suspected to show deterioration. A measure of this ability to consistently define change, sensitivity, was only 64% to 81%. This highlights the tricky business, at least in this study as designed with independent masked graders, of reliably and robustly detecting structural change over time.
These changes as studied, are detected 'in a vacuum', without knowledge of any other clinical factors, something that would not (and should not) be done in ordinary clinical practice. Quantitative measurements made over time as with scanning laser ophthalmoscopy, optical coherence tomography, or scanning laser polarimetry, may offer some help in the future. However, validated techniques with which to detect change, even with these techniques, are not yet available.