Purchase this article with an account.
Parham Azarbod, Dennis Mock, Elena Bitrian, Afifi A. Abdelmonem, Fei Yu, Kouros Nouri-Mahdavi, Anne Coleman, Joseph Caprioli; Validation Of A Novel Method For Visual Field (VF) Decay Rate Analysis. Invest. Ophthalmol. Vis. Sci. 2012;53(14):2262.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To investigate the accuracy of Pointwise Exponential Regression (PER) of thresholds to predict future global indices (MD and VFI) versus linear regression of the indices themselves.
Open-angle glaucoma patients with reliable visual fields who had 6 or more years of follow up and at least 8 visual fields were included. Pointwise forecasts were made based on a recently described model. Final threshold values (in dB) were predicted for 54 test locations based on performing a PER of the first four years or half (if total follow up>8 years) of the available data. MD and VFI forecasts were calculated by 1) extrapolation of linear regression of MD and VFI, and 2) calculation de novo from the predicted final threshold values obtained with PER. The predicted MD and VFI derived with the two methods were compared with the mean of the actual values at the end of follow-up.
Six Hundred and nine patients (798 eyes) were analyzed. The average follow up (+SD) was 8.7 (+ 2.2) years, the number of VF was 15.2 (+4.9) and MD was -8.8 (+ 6.4). The following Table compares the predicted and actual values obtained with linear regression of indices and calculation of the indices de novo after PER, and shows the Mean Squared Error (MSE), the Average of Absolute Difference (AD) in dB for MD and % for VFI, and the Area Under the Curve (AUC) of the frequency distribution (binary classifier) curves where the number of eyes vs AD is plotted.Analysis of the AUC’s shows that the PER model is significantly (p<0.001) more accurate in predicting both the MD and the VFI.
The VFI and MD can be predicted more accurately by PER of individual thresholds rather than simple linear projections of these indices. This method is valid across a wide range of VF severities and has the added advantages of preserving and presenting spatial VF information with no discrimination for a particular VF test strategy or requirement for a normative data set.
This PDF is available to Subscribers Only