Purchase this article with an account.
J. P. Pascual, J. Paetzold, U. Schiefer, L. Zangwill, R. N. Weinreb, P. A. Sample; Determining Glaucomatous Progression by Linear Regression of Visual Field Locations Within Regions in the Diagnostic Innovations in Glaucoma Study (DIGS). Invest. Ophthalmol. Vis. Sci. 2008;49(13):1092. doi: https://doi.org/.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Linear regression for groups of visual field (VF) locations is compared to regressions of single locations followed across time.
VF progression was analyzed in DIGS participants with at least 4 reliable, full threshold, standard automated perimetry tests and a stereophotograph history indicative of progressive glaucomatous optic neuropathy (pGON). To assess spatial agreement between linear regressions of groups of VF locations and single locations, a circular window of 4.3 degrees radius was convolved across the VF to create 184 regions that each contained more than 1 test location, with the centers of the regions located between the vertices of the pattern 24-2 test grid. Pattern deviation (PD) plot values for regions and locations were regressed, and progression was defined as a regression slope less than -0.5 dB/year, p<0.05. The time to endpoints for progression in single locations and for various numbers of progressed regions containing that location was compared. Survival curves characterized the time to endpoint. Sensitivity and specificity for VF progression for various numbers of locations at endpoint were determined using several cutoffs for number of regions at endpoint.
The number of VF exams from the 37 eligible pGON patients was 15.8 ± 6.0 (mean ± SD) exams over 11.5 ± 3.6 years with an interval of 0.78 ± 0.73 years between exams. MD and PSD at the first exam were -1.09 ± 2.14 dB and 3.26 ± 2.53 dB, respectively. The changes in MD and PSD between the first and last exams are -3.06 ± 3.45 dB and 1.42 ± 2.46 dB. The median survival time for a location to reach endpoint was 11.8, 4.6, 2.0, 1.0 and 0.4 years when 1, 2, 3, 4, and 5 regions progressed. The median survival time is the time at which half of locations reach their endpoint after their regions reach their endpoints. Sensitivities for 4-location progression are 0.80 and 0.70 and specificities are 0.88 and 0.94 when 12 and 13 regions in the VF are at endpoint (out of 184 possible regions).
The region-based algorithm in this study is comparable to traditional location-based regression techniques, and location endpoints coincided with region progression endpoints. This region-based progression algorithm would be applicable to a new generation of VF exams that employ condensed test grids since it functions independent of the test grid’s spatial arrangement.
Clinical Trial: :
University of Sao Paulo #310/05
This PDF is available to Subscribers Only