Purchase this article with an account.
Chris Johnson, Carrie Doyle, Trina Eden, Michael Wall; Temporal filtering of longitudinal visual field data. Invest. Ophthalmol. Vis. Sci. 2013;54(15):3931.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To determine whether low pass filtering (3 temporal bin smoothing) of longitudinal visual field data would influence the ability to detect progressive glaucomatous damage.
One eye of one hundred and eight patients with a clinical diagnosis of glaucoma (MD mean at baseline = -6.73, SD = 4.45, Max = -0.12, Min = -19.94) underwent automated perimetric testing with the Humphrey Field Analyzer 24-2 stimulus pattern, a size III target and the SITA standard threshold estimation strategy. Testing was performed every six months, with two initial baseline visual field measures. To be eligible for the study, participants were required to have at least five additional visual field tests beyond the initial baseline measurements. Mean Deviation (MD) and Pattern Standard Deviation (PSD) are often used as clinical methods of monitoring glaucomatous visual field progression, as is the visual field index or VFI which is based on MD. Linear regression was used to assess the filtered and unfiltered data sets.
We found minimal differences in the slope of linear regression results for filtered versus unfiltered MD (Unfiltered mean = -0.446, SD = 0.703, max = 0.869, min = -2.45; Filtered mean = -0.316, SD = 0.597, Max = 0.919, Min = -2.17) or PSD (Unfiltered mean = 0.102, SD = 0.410, Max = 2.25, Min = -0.958; Filtered mean = 0.044, SD = 0.436, Max = 2.099, Min = -0.977) or in the goodness of fit for the regression evaluations.
Although previous investigations of procedures to minimize variability for longitudinal visual field data sets (e.g. low pass filtering and one-omitting and three-omitting analysis methods) have reported modest improvements in the ability to determine glaucomatous visual field changes, our findings revealed no difference between filtered and unfiltered data sets in assessing visual field change. This may be due to our use of global indices (MD and PSD) that provide a composite evaluation of all visual field locations tested (reducing variability) rather than individual points or small groups of points evaluated in prior studies, and the use of empirical visual field data rather than simulated results.
This PDF is available to Subscribers Only