Criteria for progression of glaucomatous visual field loss considering slope and scatter of local serial differential luminance sensitivity data

Abstract

Abstract: : Purpose: To improve evaluation of progression of glaucomatous visual field defects (VFDs) by considering scatter in addition to slope of local differential luminance sensitivity (DLS) threshold values over time. Methods: 113 eyes of 87 glaucoma patients were examined with a 4–2–1 dB thresholding algorithm using the Tuebingen Computer Campimeter (TCC) within 30° eccentricity every 6 to 9 months. Local DLS threshold was calculated by maximum–likelihood estimation. Consensus readings of 49 patients’ local and total "loss volumes" on at least 3 examinations illustrated by graphical output were the basis for our first discriminant analysis. Expert opinion on 285 points in the visual fields of 8 subjects after visual inspection of printed visual field series of 6 to 7 examinations were combined with slope and scatter estimates of pointwise linear regressions (PLR) by discriminant analysis in a second experiment. The procedure was initially applied to 81 randomly chosen locations of 4 patients and then repeated with 211 locations of 5 patients, which were selected to be difficult to classify according to the first results. The optimal rule was required to be symmetric about the line of no slope and compared among data sets to establish repeatability. Results: A negative slope exceeding –1.38 dB per year was worked out as a criterion for progression in PLR, as long as the residual standard deviation (SD) was less than 2.41 dB + 1/5 year ' slope (dB / year). A SD of 2.69 dB or less enables one to call flatter slopes stable disease. Classification of the human expert and the above rule coincided in 79% of all points, resulting in coefficient of agreement kappa of 0.63. Rules estimated from both data sets coincided in 90% of 4652 points (kappa 0.86). Similar results were established for local and total defect volume. Conclusions: High variability may be an early sign of glaucoma, but precludes extrapolation of a trend. The relevance of that trend can be judged by our simple rule that consideres scatter also and was derived in a repeatable manner.

Keywords: perimetry • neuro–ophthalmology: diagnosis • clinical research methodology